WO2024161469A1 - Traffic signal recognition method and traffic signal recognition device - Google Patents
Traffic signal recognition method and traffic signal recognition device Download PDFInfo
- Publication number
- WO2024161469A1 WO2024161469A1 PCT/JP2023/002897 JP2023002897W WO2024161469A1 WO 2024161469 A1 WO2024161469 A1 WO 2024161469A1 JP 2023002897 W JP2023002897 W JP 2023002897W WO 2024161469 A1 WO2024161469 A1 WO 2024161469A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- result
- judgment
- traffic light
- sequence
- interlocking
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Ceased
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
-
- G—PHYSICS
- G08—SIGNALLING
- G08G—TRAFFIC CONTROL SYSTEMS
- G08G1/00—Traffic control systems for road vehicles
- G08G1/09—Arrangements for giving variable traffic instructions
Definitions
- the present invention relates to a traffic light recognition method and a traffic light recognition device.
- Patent Document 1 proposes a traffic light recognition technology that accumulates the traffic light color (current display state) for each frame across multiple frames, and performs a majority vote on the traffic light color results for multiple frames to determine the latest traffic light color.
- Patent Document 1 involves majority voting on the results of the current signal status of a traffic light in multiple frames, which creates the problem of a delay between when the current signal status of a traffic light changes and when it is actually determined that the current signal status has changed.
- the present invention was made in consideration of the above problems, and its purpose is to provide a traffic light recognition method and a traffic light recognition device that can reduce the delay between when the current signal state of a traffic light changes and when it is actually determined that the current signal state has changed.
- a traffic light recognition method and traffic light recognition device targets multiple images captured in the direction of vehicle travel, determines the current status of each traffic light for each image, and obtains a result sequence consisting of multiple determination results in chronological order for each traffic light. If there are interlocking traffic lights among the multiple traffic lights, the result sequences for each interlocking traffic light are integrated across the interlocking traffic lights to generate an integrated sequence, and the current status associated with the most frequent determination result among the determination results constituting the integrated sequence is output as the current status based on the traffic light.
- the present invention makes it possible to reduce the delay between when the current signal state of a traffic light changes and when it is actually determined that the current signal state has changed.
- FIG. 1 is a block diagram showing the configuration of a traffic light recognition device according to an embodiment of the present invention.
- FIG. 2 is a flowchart showing a processing procedure of the traffic light recognition device according to one embodiment of the present invention.
- FIG. 3 is a flowchart showing the output determination process.
- FIG. 4 is a diagram showing an example of a result sequence made up of a plurality of determination results in chronological order.
- FIG. 5 shows an example of a current status output based on a result sequence for one traffic light.
- FIG. 6A is a diagram showing an example of integrating two result strings.
- FIG. 6B is a diagram showing an example of integrating three result strings.
- FIG. 7 is a diagram showing an example of integrating two result strings taking into account the chronological order.
- FIG. 8 is a diagram showing an example of the output of the current status when the number of interlocking traffic signals is reduced.
- FIG. 9 is a diagram showing an example of the output of the current status when the number of interlocking traffic signals is increased.
- FIG. 10 is a flowchart showing the process of calculating the size of a subsequence.
- FIG. 11 is a diagram showing an example of merging two result strings by varying the size of the substrings.
- FIG. 12 is a diagram showing an example of complementation of the determination results in the result string.
- Fig. 1 is a block diagram showing the configuration of a traffic light recognition device according to this embodiment.
- the traffic light recognition device according to this embodiment includes an imaging unit 71 and a controller 100.
- the traffic light recognition device may include an on-board sensor 73, a map information acquisition unit 75, and a vehicle control device 400.
- the controller 100 is connected to the imaging unit 71, the on-board sensor 73, the map information acquisition unit 75, and the vehicle control device 400 via a wired or wireless communication path.
- the imaging unit 71, the on-board sensor 73, and the vehicle control device 400 are mounted on a vehicle (not shown), but the map information acquisition unit 75 and the controller 100 may be mounted on the vehicle or installed outside the vehicle.
- the imaging unit 71 captures images in the direction in which the vehicle is traveling.
- the imaging unit 71 is a digital camera equipped with a solid-state imaging element such as a CCD (Charge Coupled Device) or a CMOS (Complementary Metal-Oxide Semiconductor), and captures images of the surroundings of the vehicle to obtain digital images of the surrounding area.
- the imaging unit 71 captures an image of a specified range around the vehicle by setting the focal length, the lens angle of view, the vertical and horizontal angles of the camera, etc.
- the captured images captured by the imaging unit 71 are output to the controller 100 and stored in a storage unit (not shown) for a predetermined period of time.
- the imaging unit 71 captures captured images at predetermined time intervals, and the captured images captured at the predetermined time intervals are stored in the storage unit as past images.
- the past images may be deleted after a predetermined period of time has passed since the capture of the past images.
- the on-board sensor 73 is composed of an object detection sensor, such as a laser radar, a millimeter wave radar, or a camera, mounted on the vehicle, that detects objects present around the vehicle.
- the on-board sensor 73 may be equipped with a plurality of different types of object detection sensors.
- the on-board sensor 73 detects the environment around the vehicle.
- the on-board sensor 73 may detect moving objects including other vehicles, motorcycles, bicycles, and pedestrians, and stationary objects including stopped vehicles, and detect the position, attitude, size, speed, acceleration, deceleration, yaw rate, etc. of the moving and stationary objects relative to the vehicle.
- the on-board sensor 73 may output, as a detection result, the behavior of two-dimensional objects in a zenith diagram (also called a plan view) viewed from the air above the vehicle.
- the on-board sensor 73 may also detect signs (road signs and signs marked on the road surface) and guide rails that exist around the vehicle.
- the on-board sensor 73 may detect the rotational speed or difference in rotational speed of the wheels of the vehicle to detect the slipperiness of the road surface on the lane on which the vehicle is traveling.
- the on-board sensor 73 detects the state of the vehicle. For example, the on-board sensor 73 may detect the vehicle's moving speed (forward/backward and left/right moving speed, turning speed), the steering angle of the wheels of the vehicle, and the speed of change of the steering angle.
- the vehicle's moving speed forward/backward and left/right moving speed, turning speed
- the steering angle of the wheels of the vehicle and the speed of change of the steering angle.
- the on-board sensor 73 may include a sensor that measures the absolute position of the vehicle, that is, the position, attitude, and speed of the vehicle relative to a predetermined reference point, using a position detection sensor that measures the absolute position of the vehicle, such as a GPS (Global Positioning System) or odometry.
- a position detection sensor that measures the absolute position of the vehicle, such as a GPS (Global Positioning System) or odometry.
- the map information acquisition unit 75 acquires map information that indicates the structure of the road on which the vehicle is traveling.
- the map information acquired by the map information acquisition unit 75 includes road structure information such as absolute lane positions, lane connection relationships, and relative position relationships.
- the map information acquired by the map information acquisition unit 75 may also include facility information such as parking lots and gas stations.
- the map information may include traffic light position information, traffic light type, and the position of the stop line corresponding to the traffic light.
- the map information acquisition unit 75 may own a map database that stores map information, or may acquire map information from an external map data server by cloud computing.
- the map information acquisition unit 75 may also acquire map information using vehicle-to-vehicle communication or road-to-vehicle communication.
- the vehicle control device 400 controls a vehicle (not shown) based on the traffic light recognition results obtained by the controller 100.
- the vehicle control device 400 may be a device that automatically drives the vehicle along a predetermined driving route, or a device that assists the vehicle occupant in driving operations.
- the vehicle control device 400 may be a notification device that notifies the vehicle occupant of the traffic light recognition results.
- the controller 100 (an example of a control unit or processing unit) is a general-purpose microcomputer equipped with a CPU (central processing unit), memory, and input/output units.
- a computer program (traffic light recognition program) for functioning as part of the traffic light recognition device is installed in the controller 100. By executing the computer program, the controller 100 functions as multiple information processing circuits (140, 150, 160, 165, 170, 180) equipped in the traffic light recognition device.
- the multiple information processing circuits (140, 150, 160, 165, 170, 180) of the traffic light recognition device are realized by software.
- the multiple information processing circuits (140, 150, 160, 165, 170, 180) may be configured by individual hardware.
- the information processing circuits (140, 150, 160, 165, 170, 180) may also be used as electronic control units (ECUs) used for other vehicle-related controls.
- ECUs electronice control units
- the controller 100 includes a detection area setting unit 140, a judgment unit 150, a judgment result storage unit 160, an integration unit 165, an output determination unit 170, and an output unit 180 as multiple information processing circuits (140, 150, 160, 165, 170, 180).
- the detection area setting unit 140 sets a detection area corresponding to a traffic light on the image captured by the imaging unit 71.
- detection area means an area in the image where a traffic light is estimated to be present.
- the detection area setting unit 140 sets a detection area for each traffic light, for one or more traffic lights estimated to be present in the image.
- the position of the traffic light captured within the imaging range in the image can be estimated based on the imaging direction of the imaging unit 71, the position and posture of the vehicle at the time of imaging, and the position of the traffic light.
- the detection area setting unit 140 sets, for example, a part of the captured image that includes the estimated position of the traffic light in the image as the detection area.
- the detection area setting unit 140 may extract, from among the traffic lights captured within the imaging range, those that correspond to the road on which the vehicle is traveling as target traffic lights, and set a detection area for each target traffic light, targeting only the extracted target traffic lights. For example, the detection area setting unit 140 extracts, as the target traffic light, a traffic light that corresponds to the road on which the vehicle is traveling, based on map information that indicates the position of the vehicle and the structure of the road on which the vehicle is traveling. This makes it possible to output the current indication status of the target traffic light based on the result sequence related to the target traffic light or the integrated sequence, through processing by the determination unit 150, determination result storage unit 160, integration unit 165, output determination unit 170, and output unit 180, which will be described later.
- the detection area setting unit 140 also determines whether or not there is an "interlocking traffic light” among the target traffic lights.
- "interlocking traffic lights” refers to multiple traffic lights that correspond to the same stop line.
- the detection area setting unit 140 may directly obtain the position of the stop line corresponding to the traffic lights from the map information, and determine whether or not multiple traffic lights are "multiple traffic lights” based on whether or not the positions of the stop lines are the same.
- the detection area setting unit 140 may also determine whether or not multiple traffic lights are "multiple traffic lights” based on position information of the traffic lights.
- the detection area setting unit 140 may also determine whether or not multiple traffic lights are "multiple traffic lights” based on position information of the multiple traffic lights detected by the in-vehicle device.
- the determination unit 150 performs image processing on the detection area set for each traffic light, detects the traffic light within the detection area, and determines the current status of the traffic light.
- the determination unit 150 detects traffic lights by, for example, template matching. Template matching uses an image of a standard traffic light as a template, scans the detection area while shifting it by one pixel at a time, and calculates the correlation of the brightness distribution, for example. Then, when the correlation reaches the highest value, it detects that a traffic light is present at the position on the image where the template is located.
- the "color signals” shown by traffic lights include “green light,” “yellow light,” and “red light.”
- the meaning of a "color signal” is determined by traffic laws that vehicles must follow; for example, a “green light” means “you may proceed,” a “red light” means “stop at the stop position,” and a “yellow light” means “stop at the stop position, unless you are too close to the stop position to stop safely.”
- a “green light” means “permission to start” for a vehicle
- a “red light” and a “yellow light” have meanings other than “permission to start” for a vehicle.
- Such discrimination between “green light,” “yellow light,” and “red light” may be performed by assuming that the "color signal” with the highest brightness level among the three “color signals” is illuminated.
- traffic lights are not limited to "color signals,” but may also display “arrow signals” that indicate the direction vehicles are permitted to go at the intersection where the traffic lights are installed. Examples include “right turn signals,” “straight ahead signals,” and “left turn signals.”
- Arrow signals are not limited to “right turn signals,” “straight ahead signals,” and “left turn signals,” and there are many variations depending on the structure of the intersection where the traffic lights are installed. The meaning of an “arrow signal” is determined by the traffic laws that vehicles must follow.
- the determination unit 150 performs image processing on the detection area to determine the current status of the traffic light, such as the illumination status of the "color signal” or "arrow signal.”
- the image processing for traffic light detection in the determination unit 150 may utilize machine learning such as a support vector machine or a neural network.
- machine learning such as a support vector machine or a neural network.
- a learning database in which templates of traffic lights of different sizes are stored in advance can be prepared, and the recognition rate can be improved by referring to different learning databases depending on the distance to the traffic light.
- the determination result storage unit 160 stores the current state of one or more traffic lights determined by the determination unit 150 for each traffic light. Specifically, the determination unit 150 sequentially performs a determination for each traffic light on multiple images in chronological order acquired by the imaging unit 71, and the determination result storage unit 160 acquires the obtained determination results from the determination unit 150. The determination result storage unit 160 then stores the multiple determination results in chronological order for each traffic light as a result sequence associated with the traffic light.
- Figure 4 is a diagram showing an example of a result string for one traffic light, consisting of multiple judgment results in chronological order.
- the judgment results obtained by the judgment unit 150 are assumed to be of four types: "green”, “yellow”, “red”, and “unknown” (a judgment result in which the current display status is unknown).
- the judgment results of "green”, “yellow”, “red”, and “unknown” are represented by a square filled with white, a square filled with a diagonal stripe pattern, a square filled with a checkerboard pattern, and a square marked with a "?” mark, respectively. The same notation is used in the subsequent figures.
- the reason why the determination unit 150 may determine the status as "unknown” is, for example, when the traffic light flickers due to a flicker phenomenon (when the traffic light is an LED lamp that flashes periodically, the brightness in the captured image changes depending on the timing of the image capture by the image capture unit 71) or a rolling shutter phenomenon (when the image capture method used by the image capture unit 71 causes the image capture timing to differ for each position in the captured image).
- a flicker phenomenon when the traffic light is an LED lamp that flashes periodically, the brightness in the captured image changes depending on the timing of the image capture by the image capture unit 71
- a rolling shutter phenomenon when the image capture method used by the image capture unit 71 causes the image capture timing to differ for each position in the captured image.
- the number N of judgment results that constitute one result sequence associated with one traffic light and are stored by the judgment result storage unit 160 is preset.
- the N judgment results that make up one result string associated with one traffic light are each assigned a subscript ranging from 0 to N-1.
- the judgment result with subscript 0 is the most recent judgment result obtained by the judgment unit 150, and as the subscript assigned to the judgment result increases, it indicates that the judgment result is older.
- the judgment result storage unit 160 increments the subscript assigned to the judgment result by 1 and assigns the subscript 0 to the new judgment result input to the judgment result storage unit 160.
- the judgment result with the subscript N is erased.
- the judgment results with subscripts from 0 to N-1 are stored in the judgment result storage unit 160 as a result string.
- the judgment result storage unit 160 may have a function of supplementing the judgment results. Specifically, when "unknown” is included in the judgment results constituting the result sequence stored in the judgment result storage unit 160, the judgment result of "unknown” may be replaced based on the judgment result immediately before the "unknown” and the judgment result immediately after the "unknown".
- the judgment result for subscript 2 is "unknown.”
- the judgment result storage unit 160 refers to the judgment result for subscript 3, which is immediately before the judgment result for subscript 2, and the judgment result for subscript 1, which is immediately after the judgment result for subscript 2.
- the judgment results for subscript 1 and subscript 3 are both "yellow light.” Therefore, the judgment result for subscript 2 may be "yellow light,” which is the judgment result for subscript 3, which is immediately before the judgment result for subscript 2, rather than "unknown.”
- the judgment result storage unit 160 may extract, from among the judgment results constituting the result string, a judgment result that indicates that the current display state is unknown as the first judgment result.
- the judgment result storage unit 160 may then extract, from among the judgment results constituting the result string, excluding judgment results that indicate that the current display state is unknown, the judgment result immediately before the first judgment result and the judgment result immediately after the first judgment result as the second judgment result and the third judgment result, respectively, and may replace the first judgment result in the result string with the second judgment result when the second judgment result and the third judgment result are identical.
- the judgment result storage unit 160 may replace the first judgment result in the result string with the judgment result immediately following the first judgment result if there is no judgment result immediately preceding the first judgment result among the judgment results constituting the result string, excluding the first judgment result. This makes it possible to complement the judgment results even if there is a judgment result at the beginning of the result string that indicates that the current display state is unknown.
- the judgment result storage unit 160 may replace the first judgment result in the result string with the judgment result immediately preceding the first judgment result if there is no judgment result immediately following the first judgment result among the judgment results constituting the result string, excluding the first judgment result. This makes it possible to complement the judgment results even if there is a judgment result at the end of the result string indicating that the current display state is unknown.
- the function of complementing the determination result by the determination result storage unit 160 described above is effective in improving the accuracy of traffic light recognition.
- the determination result storage unit 160 may determine whether or not an interlocking traffic signal has changed from a state in which there is an interlocking traffic signal to a state in which there is no interlocking traffic signal. If it is determined that an interlocking traffic signal has changed from a state in which there is an interlocking traffic signal to a state in which there is no interlocking traffic signal, the determination result storage unit 160 may replace the determination results constituting the result sequence related to the traffic signal with the determination results constituting the integrated sequence.
- the judgment result storage unit 160 may increase the subscript assigned to the judgment result by 1 for the result string after it has been replaced with the judgment results that make up the integrated string, and assign the subscript 0 to the new judgment result input to the judgment result storage unit 160.
- the judgment result with the subscript N is erased.
- the judgment results stored in the judgment result storage unit 160 are described as N, but this is not limited to this.
- the judgment result storage unit 160 may determine whether or not to store a judgment result based on the capture time of the image on which the judgment result is based, instead of the number of judgment results. More specifically, the judgment result storage unit 160 may erase judgment results corresponding to captured images captured a predetermined time or more after capture, so that the judgment results stored in the judgment result storage unit 160 are only those corresponding to images captured up to a predetermined time in the past from the present.
- the imaging unit 71 captures images at a predetermined cycle
- the result is the same whether the number of judgment results stored in the judgment result storage unit 160 is limited based on the number N or limited based on a predetermined time.
- the results may differ depending on whether the number is limited based on the number N or based on a specified time.
- the configuration of the judgment result storage unit 160 can be adopted regardless of whether the number of judgment results to be stored is limited based on the number N or based on a specified time.
- the output determination unit 170 determines the output value based on the judgment results included in the result column for each traffic light or the judgment results included in the integrated column described below.
- the output determination unit 170 sets the current indication state associated with the most frequent judgment result among the judgment results constituting the result string related to the traffic light as the output value.
- the output value is set to the current indication state associated with the judgment result determined by majority vote among the judgment results constituting the result string.
- the current status related to the judgment result to be used preferentially when setting the output value may be set in advance.
- the output value may be set preferentially in the following order: "red light,” “yellow light,” “green light,” and "unknown.”
- FIG. 5 is a diagram showing an example of the output of the current display state based on the result sequence related to one traffic light.
- FIG. 5 shows how the result sequence related to one traffic light transitions from step "t" to step "t+13".
- step "t” and step "t+4" the majority of the judgment results in the result sequence are related to "red light", so the output determination unit 170 sets the output value to "red”.
- step "t+5" there are the same number of judgment results related to "red light” and “green light”, and it is not possible to determine which is the most common judgment result. Therefore, the output determination unit 170 prioritizes “red light” over “green light”, and sets the output value to "red”.
- step "t+6” and step “t+11” the majority of the judgment results in the result sequence are related to "green light", so the output determination unit 170 sets the output value to "green”.
- step "t+12” and step “t+13” the majority of the judgment results in the result sequence are related to "yellow light", so the output determination unit 170 sets the output value to "yellow”.
- the output determination unit 170 sets the current indication state associated with the most frequent judgment result among the judgment results that make up the integrated sequence, which will be described later, as the output value.
- the output value is set to the current indication state associated with the judgment result determined by majority vote among the judgment results that make up the integrated sequence.
- the method of determining the output value based on the integrated sequence, performed by the output determination unit 170, is the same as the method of determining the output value based on the result sequence described above.
- the output determination unit 170 when there are no interlocking traffic lights, the output value is determined based on the result sequence, and when there are interlocking traffic lights, the output value is determined based on the integrated sequence.
- the integration unit 165 may overwrite the integrated sequence with the determination result of the result sequence. In this way, the output determination unit 170 may always determine the output value based on the integrated sequence regardless of the number of traffic lights.
- the integration unit 165 integrates the result sequences for each of the interlocking traffic signals across the interlocking traffic signals to generate an integrated sequence. For example, among the judgment results that constitute the result sequence related to one interlocking traffic signal, the integration unit 165 integrates a subsequence consisting of a predetermined number or less of consecutive judgment results, including the most recent judgment result, across the interlocking traffic signals to generate an integrated sequence.
- the integration unit 165 may set the predetermined number as the "subsequence size calculation process" to a value equal to or less than the value obtained by dividing the number of judgment results constituting a result sequence related to one linked traffic light by the number of linked traffic lights.
- Figure 6A is a diagram showing an example of merging two result strings.
- Figure 6B is a diagram showing an example of merging three result strings.
- FIG. 6A shows the transition of the result sequence related to the first traffic light and the result sequence related to the second traffic light from step “t” to step “t+6".
- FIG. 6B further shows the transition of the result sequence related to the third traffic light from step “t" to step "t+6".
- the integration unit 165 shows an example in which the predetermined number is set to 5 (a value obtained by dividing the number of judgment results constituting the result string by 2, which is the number of traffic lights). That is, in FIG. 6A, the substring in each step is composed of five judgment results with subscripts 0 to 4.
- the integration unit 165 generates an integrated sequence by simply concatenating a subsequence that is a part of the result sequence related to the first traffic light with a subsequence that is a part of the result sequence related to the second traffic light.
- the judgment results of the subscripts 0 to 4 of the subsequence of the first traffic light are respectively used as the judgment results of the subscripts 0 to 4 of the integrated sequence.
- the judgment results of the subscripts 0 to 4 of the subsequence of the second traffic light are respectively used as the judgment results of the subscripts 5 to 9 of the integrated sequence.
- the majority decision result from step “t” to step “t+4" will be “green,” and the majority decision result from step “t+5" onwards will be “yellow.” The same applies to the majority decision result based on the result sequence related to the second traffic light.
- step "t” the majority result from step "t” to step “t+2" will be “green,” and the majority result from step “t+3" onwards will be “yellow.” Therefore, compared to when the judgment is made based on the result sequence, when the judgment is made based on the integrated sequence, the delay from when the current signal state of the traffic light changes to when it is actually determined that the current signal state has changed is reduced by two steps.
- the effect of reducing the delay between when the current signal status of a traffic light changes and when it is actually determined that the current signal status has changed becomes greater the more traffic lights that contribute to the generation of the integrated sequence.
- the integration unit 165 shows an example in which the predetermined number is set to 3 (the largest integer that does not exceed the value obtained by dividing the number of judgment results that make up the result sequence by 3, which is the number of traffic lights). That is, in FIG. 6B, the subsequence in each step is made up of three judgment results with subscripts 0 to 2.
- the integration unit 165 generates an integrated sequence by directly concatenating a subsequence that is a part of the result sequence related to the first traffic light, a subsequence that is a part of the result sequence related to the second traffic light, and a subsequence that is a part of the result sequence related to the third traffic light.
- the judgment results of subscripts 0 to 2 of the subsequence of the first traffic light are respectively used as the judgment results of subscripts 0 to 2 of the integrated sequence.
- the judgment results of subscripts 0 to 2 of the subsequence of the second traffic light are respectively used as the judgment results of subscripts 3 to 5 of the integrated sequence.
- the judgment results of subscripts 0 to 2 of the subsequence of the third traffic light are respectively used as the judgment results of subscripts 6 to 8 of the integrated sequence.
- the output determination unit 170 makes a judgment based on the integrated sequence, the majority result from step “t” to step “t+1” will be “green,” and the majority result from step “t+2" onwards will be “yellow.” Therefore, compared to when the judgment is made based on the result sequence, when the judgment is made based on the integrated sequence, the delay from when the current signal state of the traffic light changes to when it is actually determined that the current signal state has changed is reduced by three steps.
- the integration unit 165 may integrate the result sequences as they are to generate an integrated sequence.
- the integrated sequence will be made up of determination results with subscripts 0 to 19.
- the integration unit 165 may generate an integrated sequence by arranging the determination results that constitute the integrated sequence in chronological order based on the chronological order of the determination results that constitute the result sequence related to the linked traffic signals.
- FIG. 7 is a diagram showing an example of integrating two result sequences taking into account the chronological order.
- the integration unit 165 generates an integrated sequence by combining a subsequence that is a part of the result sequence related to the first traffic light and a subsequence that is a part of the result sequence related to the second traffic light, taking into account the chronological order.
- the determination results of the subscripts 0, 1, 2, 3, and 4 of the subsequence of the first traffic light are respectively set as the determination results of the subscripts 0, 2, 4, 6, and 8 of the integrated sequence.
- the determination results of the subscripts 0, 1, 2, 3, and 4 of the subsequence of the second traffic light are respectively set as the determination results of the subscripts 1, 3, 5, 7, and 9 of the integrated sequence.
- the output determination unit 170 can determine the output value based on the chronological order information in addition to the majority vote result. Furthermore, when replacing the determination results constituting the result sequence related to traffic lights with the determination results constituting the integrated sequence, it is possible to eliminate a situation in which the chronological order is lost in the determination results in the result sequence after replacement.
- Figure 8 is a diagram showing an example of the current status output when the number of linked traffic lights is reduced.
- Figure 8 shows how the result sequence related to the first traffic light transitions from step “t” to step “t+6". Also, it shows how the result sequence related to the second traffic light transitions from step “t” to step “t+3".
- step "t+4" the second traffic light disappears from the image, and only the first traffic light is visible on the image.
- the integration unit 165 shows an example in which the predetermined number is set to 5 (a value obtained by dividing the number of judgment results constituting the result string by 2, which is the number of traffic lights). That is, in FIG. 8, the substring in each step is composed of five judgment results with subscripts 0 to 4.
- step "t+4" the output decision unit 170 returns from majority decision based on the integrated sequence to majority decision based on the result sequence of the first traffic light.
- step "t+3” the majority judgment based on the integrated sequence is "yellow”
- step "t+4" the majority judgment based on the result sequence for the first traffic light is "green”.
- the judgment result of the current state of the traffic light changes frequently, such as "green” -> "yellow” -> "green”.
- the judgment results constituting the result sequence for the first traffic light are replaced with the judgment results constituting the integrated sequence, the above problem does not occur.
- the judgment results of the result sequence in step "t+3” are replaced with the judgment results of the integrated sequence in step "t+3".
- the judgment results of the subscripts 0 to 9 of the integrated sequence are respectively set as the judgment results of the subscripts 0 to 9 of the result sequence for the first traffic light.
- step "t+4" the majority decision based on the result sequence of the first traffic light can be "yellow". This makes it possible to prevent the judgment results of the current state of the traffic light from changing frequently in a very short step.
- the integrating unit 165 may determine whether or not the number of interlocking traffic lights has increased since the previous generation of the integrated sequence. When it is determined that the number of interlocking traffic lights has increased, the integrating unit 165 may generate an integrated sequence by integrating the result sequences for each interlocking traffic light other than the newly added interlocking traffic light until the number of determination results constituting the result sequence related to the newly added interlocking traffic light reaches or exceeds a predetermined threshold.
- the integration unit 165 may set the predetermined threshold to a value equal to or smaller than the value obtained by dividing the number of judgment results constituting the result sequence related to one traffic light by the number of traffic lights.
- FIG. 9 is a diagram showing an example of the output of the current status when the number of interlocking traffic lights is increased.
- FIG. 9 shows how the result sequence related to the first traffic light and the result sequence related to the second traffic light transition from step "t" to step "t+6".
- the second traffic light which was not captured in the image in steps prior to step "t"
- the judgment results that make up the result sequence related to the second traffic light with the subscripts 1 to 9 excluding the judgment result with the subscript 0, are "unknown”.
- the integration unit 165 shows an example in which the predetermined threshold is set to 5 (a value obtained by dividing the number of judgment results that make up the result sequence by 2, which is the number of traffic lights).
- the integrating unit 165 does not use the result sequence related to the second traffic light in generating the integrated sequence until the number of judgment results that make up the result sequence related to the second traffic light becomes equal to or greater than a predetermined threshold.
- a predetermined threshold In the example shown in FIG. 9, at the time of step "t+4", the number of judgment results that make up the result sequence related to the second traffic light becomes 5, which is equal to or greater than the predetermined threshold. Therefore, the integrating unit 165 does not use the result sequence related to the second traffic light in generating the integrated sequence from step "t” to step "t+3". On the other hand, the integrating unit 165 uses the result sequence related to the second traffic light in generating the integrated sequence from step "t+4" onwards.
- the integration unit 165 may not use the result sequence for the newly added traffic light to generate the integrated sequence until a sufficient number of judgment results are accumulated in the result sequence for the newly added traffic light. This prevents a decrease in the accuracy of the majority decision based on the integrated sequence.
- the output unit 180 outputs the output value determined by the output determination unit 170 as the "current indication state based on the traffic light.”
- the "current indication state based on the traffic light” can be said to be the result obtained by integrating the current indication states of one or more traffic lights.
- the "current status based on the traffic light” is output from the output unit 180 to the vehicle control device 400 and used to control the vehicle.
- the "current status based on the traffic light” may be output from the output unit 180 to a reporting device (not shown) and notified to the occupant via the reporting device.
- FIG. 2 is a flowchart showing the process by the traffic light recognition device according to the present embodiment.
- Figure 3 is a flowchart showing an output determination process.
- the traffic light recognition process shown in Figures 2 and 3 may be executed each time an image is captured by the imaging unit 71, or may be executed at each cycle in which image processing is performed on the detection area after an image is captured by the imaging unit 71.
- step S101 the controller 100 acquires an image of one or more traffic lights in the direction in which the vehicle is traveling.
- step S103 the controller 100 acquires the vehicle's position via the on-board sensor 73, etc.
- step S105 the controller 100 acquires the positions of one or more traffic lights based on the map information acquired from the map information acquisition unit 75 or the like.
- step S107 the detection area setting unit 140 extracts the traffic light corresponding to the road on which the vehicle is traveling as the target traffic light. Note that if all traffic lights reflected in the acquired image are to be treated as target traffic lights, the processing of steps S103, S105, and S107 may be omitted.
- step S109 the determination result storage unit 160 determines whether the state has changed from one in which there is an interlocking traffic light to one in which there is not.
- step S111 the determination result storage unit 160 replaces the result sequence related to the traffic light with the integrated sequence.
- the determination result storage unit 160 replaces the determination results constituting the result sequence related to the traffic light with the determination results constituting the integrated sequence.
- step S111 is skipped and the process proceeds to step S113.
- step S113 the detection area setting unit 140 sets a detection area for each traffic light on the image.
- step S115 the determination unit 150 performs image processing on the detection area set for each traffic light to detect the traffic lights within the detection area, and determines the current status for each traffic light.
- step S117 the determination result storage unit 160 stores the result sequence for each traffic light.
- step S119 the output determination unit 170 performs the output determination process shown in FIG. 3.
- step S121 the output determination unit 170 determines whether or not there is a linked traffic light.
- step S123 the output determination unit 170 performs a "subsequence size calculation process.”
- step S125 the integration unit 165 generates an integrated sequence based on the set subsequence.
- step S127 the output determination unit 170 determines the output value based on the integrated column.
- step S131 the output determination unit 170 determines the output value based on the result sequence of one traffic light.
- step S129 the output determination unit 170 outputs the output value to the output unit 180.
- the output unit 180 outputs the output value determined by the output determination unit 170 as the "current indication state based on the traffic light.”
- the size of the subsequence (the number of judgment results that make up the subsequence) has been described as being common among multiple traffic lights. Also, the size of the subsequence has been described as being determined based on the number of judgment results that make up the result sequence of the traffic light, and is invariant.
- the size of the subsequence may be varied based on the latest judgment result in the traffic light result sequence.
- the integration unit 165 may integrate a subsequence consisting of consecutive judgment results, including the latest judgment result, among the judgment results constituting the result sequence related to one interlocking traffic light across the interlocking traffic lights to generate an integrated sequence.
- the number of judgment results constituting a subsequence whose latest judgment result indicates something other than permission to start the vehicle may be smaller than the number of judgment results constituting a subsequence whose latest judgment result indicates permission to start the vehicle.
- FIG. 10 is a flowchart showing the process of calculating the size of a subsequence.
- the integration unit 165 selects one result sequence from among the result sequences of unselected traffic lights.
- step S203 the integration unit 165 determines whether the most recent judgment result (the judgment result with the subscript 0) among the judgment results that make up the selected result string is "green.” In other words, the integration unit 165 determines whether it indicates permission to start the vehicle.
- step S205 the integration unit 165 sets the first number as the size of the subsequence related to the selected result sequence.
- step S207 the integration unit 165 sets a second number as the size of the subsequence related to the selected result sequence.
- the second number is smaller than the first number.
- step S209 the integration unit 165 determines whether there are any unselected result columns.
- step S209 If there is an unselected result string (YES in step S209), return to step S201. If there is no unselected result string (NO in step S209), end the substring size calculation process.
- the size of the subsequence is set to the first number and the second number for the cases where the latest judgment result is "blue” and where the latest judgment result is not “blue”, but this is not limiting.
- the first number, second number, and third number may be set to correspond to the cases where the latest judgment result is "blue", “yellow”, and “red”, respectively.
- the second number may be smaller than the first number
- the third number may be smaller than the second number.
- FIG. 11 shows an example of integrating two result sequences by varying the size of the subsequences.
- FIG. 11 shows how the result sequence related to the first traffic light and the result sequence related to the second traffic light transition from step "t" to step "t+6.”
- step "t” the most recent judgment result is "green”, so the size of the subsequence is set to 5.
- step "t+1" and step "t+6 the most recent judgment result is "yellow”, so the size of the subsequence is set to 3.
- step "t” and step "t+1” the most recent judgment result is “green”, so the size of the subsequence is set to 5.
- step "t+2" to step “t+6” the most recent judgment result is "yellow”, so the size of the subsequence is set to 3.
- the size of the integrated column (the number of judgment results that make up the integrated column) is 10 at step "t", 8 at step “t+1", and 6 from step "t+2" to step "t+6".
- the output determination unit 170 makes a decision based on the result sequence related to the first traffic light
- the majority result from step “t” to step “t+4" will be “green”
- the majority result from step "t+5" onwards will be “yellow”.
- the output determination unit 170 makes a decision based on the result sequence related to the second traffic light
- the majority result from step “t” to step “t+5" will be “green”
- the majority result from step "t+6" onwards will be "yellow”.
- the output determination unit 170 makes a judgment based on the integrated sequence generated as described above, the majority result from step “t” to step “t+2” will be “green,” and the majority result from step “t+3” onwards will be “yellow.” Therefore, compared to when the judgment is made based on the result sequence, when the judgment is made based on the integrated sequence, the delay from when the current signal state of the traffic light changes to when it is actually determined that the current signal state has changed is reduced by three steps.
- the number of judgment results constituting the subsequence whose latest judgment result indicates something other than permission to start the vehicle is set to be smaller than the number of judgment results constituting the subsequence whose latest judgment result indicates permission to start the vehicle.
- the delay until a judgment other than permission to start the vehicle is made can be more effectively suppressed.
- the traffic light recognition method and traffic light recognition device target a plurality of images captured in the direction of travel of the vehicle, determine the current status of each traffic light for each image, and obtain a result sequence consisting of a plurality of determination results in chronological order for each traffic light. If there are interlocking traffic lights among the plurality of traffic lights, the result sequences for each interlocking traffic light are integrated across the interlocking traffic lights to generate an integrated sequence, and the current status associated with the most frequent determination result among the determination results constituting the integrated sequence is output as the current status based on the traffic light.
- the traffic light recognition method and traffic light recognition device may also generate an integrated sequence by integrating a subsequence of a predetermined number or less of consecutive judgment results, including the most recent judgment result, among the judgment results that make up the result sequence related to one linked traffic light, across the linked traffic lights when there are linked traffic lights. This allows for more accurate determination of the current indication status based on the judgment results of the current indication status of multiple traffic lights.
- the traffic light recognition method and traffic light recognition device may, when there are interlocking traffic lights, set the predetermined number to a value equal to or less than the value obtained by dividing the number of determination results that make up the result sequence for one interlocking traffic light by the number of interlocking traffic lights.
- the traffic light recognition method and traffic light recognition device may also generate an integrated sequence by integrating, across all of the linked traffic lights, a subsequence of consecutive judgment results, including the most recent judgment result, among the judgment results constituting the result sequence related to one linked traffic light when there are linked traffic lights.
- the number of judgment results constituting the subsequence whose most recent judgment result indicates something other than permission to start the vehicle may be smaller than the number of judgment results constituting the subsequence whose most recent judgment result indicates permission to start the vehicle.
- the delay in making decisions other than those permitting the vehicle to start is smaller than the delay in making decisions permitting the vehicle to start. As a result, it is possible to more effectively determine the current status of traffic lights that are likely to affect vehicle driving control.
- the traffic light recognition method and traffic light recognition device may determine whether the number of interlocking traffic lights has increased since the previous generation of the integrated sequence. If it is determined that the number of interlocking traffic lights has increased, the integrated sequence may be generated by integrating the result sequences of the interlocking traffic lights other than the newly added interlocking traffic light until the number of determination results constituting the result sequence for the newly added interlocking traffic light reaches or exceeds a predetermined threshold. This prevents a decrease in the accuracy of the majority decision based on the integrated sequence due to a small number of determination results constituting the result sequence for the newly added traffic light.
- the traffic light recognition method and traffic light recognition device may also generate an integrated sequence by arranging the determination results that make up the integrated sequence in chronological order based on the chronological order of the determination results that make up the result sequence related to the linked traffic lights when there are linked traffic lights. This allows for more accurate determination of the current state based on chronological order information in addition to the majority vote result. Also, when replacing the determination results that make up the result sequence related to a traffic light with the determination results that make up the integrated sequence, it is possible to eliminate a situation in which the chronological order is lost in the determination results in the result sequence after replacement.
- the traffic light recognition method and traffic light recognition device may determine whether an interlocking traffic light has changed from a state in which it is present to a state in which it is not present after the generation of the integrated sequence. If it is determined that an interlocking traffic light has changed from a state in which it is present to a state in which it is not present, the determination result constituting the result sequence related to the traffic light may be replaced with the determination result constituting the integrated sequence. This prevents the determination result of the current status of the traffic light from changing frequently in a very short step, and as a result, it becomes easier to use the determination result of the current status for vehicle control.
- the traffic light recognition method and traffic light recognition device may also be configured to output, when there is one traffic light, the current indication state associated with the most frequent judgment result among the judgment results constituting the result string related to the traffic lights as the current indication state based on the traffic light, when there is one traffic light. This allows a judgment to be made using the judgment results constituting the result string when there is one traffic light.
- the interlocking traffic lights may be multiple traffic lights that correspond to the same stop line on the road on which the vehicle is traveling. This makes it possible to determine the current status of only the traffic lights that the vehicle should obey when traveling. As a result, erroneous recognition of the current status of traffic lights can be suppressed.
- the traffic light recognition method and traffic light recognition device may extract, from among the judgment results constituting the result string, a judgment result that indicates that the current display state is unknown as the first judgment result. From among the judgment results constituting the result string, excluding the first judgment result, the judgment result immediately before the first judgment result and the judgment result immediately after the first judgment result may be extracted as the second judgment result and the third judgment result, respectively. If the second judgment result and the third judgment result are identical, the first judgment result in the result string may be replaced with the second judgment result.
- the output determination unit 170 when determining the current state of a traffic light based on the determination results that constitute a result sequence, delays in determination can be suppressed.
- the output determination unit 170 according to the result sequence before interpolation, can set the output value to "yellow” for the first time at step "t+7".
- the output determination unit 170 can set the output value to "yellow” for the first time at step "t+6". In this way, delays in determination can be suppressed.
- the traffic light recognition method and traffic light recognition device may replace the first judgment result in the result string with the judgment result immediately following the first judgment result if there is no judgment result immediately preceding the first judgment result among the judgment results constituting the result string, excluding the first judgment result. This makes it possible to complement the judgment results even if there is a judgment result at the beginning of the result string indicating that the current status is unknown.
- the traffic light recognition method and traffic light recognition device may also replace the first judgment result in the result string with the judgment result immediately preceding the first judgment result if there is no judgment result immediately following the first judgment result among the judgment results constituting the result string, excluding the first judgment result. This makes it possible to complement the judgment results even if there is a judgment result at the beginning of the result string that indicates that the current status is unknown.
- Processing circuits include programmed processors, electrical circuits, and even devices such as application specific integrated circuits (ASICs), and circuit components arranged to perform the described functions.
- ASICs application specific integrated circuits
- Imaging unit 73 Imaging unit 73 Vehicle-mounted sensor 75 Map information acquisition unit 100 Controller 140 Detection area setting unit 150 Determination unit 160 Determination result storage unit 165 Integration unit 170 Output determination unit 180 Output unit 400 Vehicle control device
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Theoretical Computer Science (AREA)
- Traffic Control Systems (AREA)
Abstract
Description
本発明は、信号機認識方法及び信号機認識装置に関する。 The present invention relates to a traffic light recognition method and a traffic light recognition device.
特許文献1によれば、複数のフレームにわたって、フレームごとの信号機の灯火色(現示状態)を蓄積し、複数のフレームでの信号機の灯火色の結果に対して多数決処理を行って、最新の信号機の灯火色を決定する信号機認識の技術が提案されている。
しかしながら、特許文献1に記載の技術によれば、複数のフレームでの信号機の現示状態の結果に対して多数決処理を行うため、信号機の現示状態が切り替わってから、現示状態が切り替わったと実際に判定されるまでの遅延が生じるという問題がある。
However, the technology described in
本発明は、上記問題に鑑みてなされたものであり、その目的とするところは、信号機の現示状態が切り替わってから、現示状態が切り替わったと実際に判定されるまでの遅延を抑制することができる信号機認識方法及び信号機認識装置を提供することにある。 The present invention was made in consideration of the above problems, and its purpose is to provide a traffic light recognition method and a traffic light recognition device that can reduce the delay between when the current signal state of a traffic light changes and when it is actually determined that the current signal state has changed.
上述した問題を解決するために、本発明の一態様に係る信号機認識方法及び信号機認識装置は、車両の進行方向を撮像した複数枚の画像を対象として、画像ごとに信号機ごとの現示状態を判定し、時系列順の複数の判定結果からなる結果列を信号機ごとに取得する。複数の信号機の中に連動する信号機がある場合に、連動する信号機ごとの結果列を、連動する信号機にわたって統合して統合列を生成し、統合列を構成する判定結果の中で最も多い判定結果に係る現示状態を、信号機に基づく現示状態として出力する。 In order to solve the above-mentioned problems, a traffic light recognition method and traffic light recognition device according to one embodiment of the present invention targets multiple images captured in the direction of vehicle travel, determines the current status of each traffic light for each image, and obtains a result sequence consisting of multiple determination results in chronological order for each traffic light. If there are interlocking traffic lights among the multiple traffic lights, the result sequences for each interlocking traffic light are integrated across the interlocking traffic lights to generate an integrated sequence, and the current status associated with the most frequent determination result among the determination results constituting the integrated sequence is output as the current status based on the traffic light.
本発明によれば、信号機の現示状態が切り替わってから、現示状態が切り替わったと実際に判定されるまでの遅延を抑制することができる。 The present invention makes it possible to reduce the delay between when the current signal state of a traffic light changes and when it is actually determined that the current signal state has changed.
次に、図面を参照して、本発明の実施の形態を詳細に説明する。説明において、同一のものには同一符号を付して重複説明を省略する。 Next, an embodiment of the present invention will be described in detail with reference to the drawings. In the description, the same parts will be given the same reference numerals and duplicate explanations will be omitted.
[信号機認識装置の構成]
図1は、本実施形態に係る信号機認識装置の構成を示すブロック図である。図1に示すように、本実施形態に係る信号機認識装置は、撮像部71と、コントローラ100とを備える。信号機認識装置は、車載センサ73、地図情報取得部75、車両制御装置400を備えるものであってもよい。コントローラ100は、有線あるいは無線の通信路によって、撮像部71、車載センサ73、地図情報取得部75、車両制御装置400と接続されている。
[Configuration of traffic light recognition device]
Fig. 1 is a block diagram showing the configuration of a traffic light recognition device according to this embodiment. As shown in Fig. 1, the traffic light recognition device according to this embodiment includes an
ここで、撮像部71、車載センサ73、車両制御装置400は、図示しない車両に搭載されるが、地図情報取得部75、コントローラ100は、車両に搭載されるものであってもよいし、車両の外部に設置されるものであってもよい。
Here, the
撮像部71は、車両の進行方向の画像を撮像する。例えば、撮像部71はCCD(Charge Coupled Device)、CMOS(Complementary Metal-Oxide Semiconductor)等の固体撮像素子を備えたデジタルカメラであり、車両の周囲を撮像して周辺領域のデジタル画像を取得する。撮像部71は、焦点距離、レンズの画角、カメラの垂直方向及び水平方向の角度などが設定されることにより、車両の周囲の所定の範囲を撮像する。
The
なお、撮像部71によって撮像された撮像画像はコントローラ100に出力され、所定の期間の間、図示しない記憶部に記憶される。例えば、撮像部71は所定の時間間隔で撮像画像を取得しており、所定の時間間隔で取得した撮像画像が、過去画像として記憶部に記憶される。過去画像は、当該過去画像の撮像時点から所定の期間を経過した後に削除されるものであってもよい。
The captured images captured by the
車載センサ73は、車両に搭載された、レーザレーダやミリ波レーダ、カメラなど、車両の周囲に存在する物体を検出する物体検出センサなどからなる。車載センサ73は、複数の異なる種類の物体検出センサを備えるものであってもよい。
The on-
車載センサ73は、車両の周囲の環境を検出する。例えば、車載センサ73は、他車両、バイク、自転車、歩行者を含む移動物体、及び停止車両を含む静止物体を検出し、移動物体及び静止物体の車両に対する位置、姿勢、大きさ、速度、加速度、減速度、ヨーレートなどを検出するものであってもよい。車載センサ73は、検出結果として、例えば車両の上方の空中から眺めた天頂図(平面図ともいう)における、2次元の物体の挙動を出力するものであってもよい。また、車載センサ73は、車両の周囲に存在する標識(道路標識や路面表示された標識)やガイドレール等を検出するものであってもよい。その他にも、車載センサ73は、車両が備える車輪の回転速度や回転速度差を検出して、車両が走行している車線の路面の滑りやすさを検出するものであってもよい。
The on-
また、車載センサ73は、車両の周囲の環境の他にも、車両の状態を検出する。例えば、車載センサ73は、車両の移動速度(前後方向、左右方向の移動速度、旋回速度)や、車両が備える車輪の転舵角、転舵角の変化速度を検出するものであってもよい。
In addition to the environment around the vehicle, the on-
その他、車載センサ73は、GPS(グローバル・ポジショニング・システム)やオドメトリなど車両の絶対位置を計測する位置検出センサなど、車両の絶対位置を計測する位置検出センサを用いて、車両の絶対位置、すなわち、所定の基準点に対する車両の位置、姿勢及び速度を計測するセンサを含んでいてもよい。
In addition, the on-
地図情報取得部75は、車両が走行する道路の構造を示す地図情報を取得する。地図情報取得部75が取得する地図情報には、車線の絶対位置、車線の接続関係、相対位置関係などの道路構造の情報が含まれる。また、地図情報取得部75が取得する地図情報には、駐車場、ガソリンスタンドなどの施設情報も含まれうる。その他、地図情報には、信号機の位置情報や、信号機の種別、信号機に対応する停止線の位置などが含まれうる。地図情報取得部75は、地図情報を格納した地図データベースを所有してもよいし、クラウドコンピューティングにより地図情報を外部の地図データサーバから取得してもよい。また、地図情報取得部75は、車車間通信、路車間通信を用いて地図情報を取得してもよい。
The map
車両制御装置400は、コントローラ100によって得られた信号機の認識結果に基づいて、図示しない車両を制御する。例えば、車両制御装置400は、所定の走行経路に従って自動運転によって車両を走行させるものであってもよいし、車両の乗員の運転操作を支援するものであってもよい。その他、車両制御装置400は、信号機の認識結果を車両の乗員に通知する通報装置であってもよい。
The
コントローラ100(制御部または処理部の一例)は、CPU(中央処理装置)、メモリ、及び入出力部を備える汎用のマイクロコンピュータである。コントローラ100には、信号機認識装置の一部として機能させるためのコンピュータプログラム(信号機認識プログラム)がインストールされている。コンピュータプログラムを実行することにより、コントローラ100は、信号機認識装置が備える複数の情報処理回路(140、150、160、165、170、180)として機能する。
The controller 100 (an example of a control unit or processing unit) is a general-purpose microcomputer equipped with a CPU (central processing unit), memory, and input/output units. A computer program (traffic light recognition program) for functioning as part of the traffic light recognition device is installed in the
なお、ここでは、ソフトウェアによって信号機認識装置が備える複数の情報処理回路(140、150、160、165、170、180)を実現する例を示す。ただし、以下に示す各情報処理を実行するための専用のハードウェアを用意して、情報処理回路(140、150、160、165、170、180)を構成することも可能である。また、複数の情報処理回路(140、150、160、165、170、180)を個別のハードウェアにより構成してもよい。更に、情報処理回路(140、150、160、165、170、180)は、車両にかかわる他の制御に用いる電子制御ユニット(ECU)と兼用してもよい。 Here, an example is shown in which the multiple information processing circuits (140, 150, 160, 165, 170, 180) of the traffic light recognition device are realized by software. However, it is also possible to configure the information processing circuits (140, 150, 160, 165, 170, 180) by preparing dedicated hardware for executing each of the information processes shown below. Also, the multiple information processing circuits (140, 150, 160, 165, 170, 180) may be configured by individual hardware. Furthermore, the information processing circuits (140, 150, 160, 165, 170, 180) may also be used as electronic control units (ECUs) used for other vehicle-related controls.
コントローラ100は、複数の情報処理回路(140、150、160、165、170、180)として、検出領域設定部140、判定部150、判定結果記憶部160、統合部165、出力決定部170、出力部180を備える。
The
検出領域設定部140は、信号機に対応する検出領域を、撮像部71で撮像した画像上に設定する。ここで、「検出領域」とは、画像内で信号機があると推定される領域を意味する。検出領域設定部140は、画像内に存在すると推定される一又は複数の信号機に対して、信号機ごとに検出領域を設定する。
The detection
撮像範囲内に写り込む信号機の画像内での位置は、撮像部71の撮像方向、撮像時の車両の位置、姿勢、信号機の位置に基づいて推定することができる。検出領域設定部140は、例えば、撮像した画像の一部領域であって、画像内での推定した信号機の位置を含む領域を、検出領域として設定する。
The position of the traffic light captured within the imaging range in the image can be estimated based on the imaging direction of the
なお、検出領域設定部140は、撮像範囲内に写り込む信号機のうち、車両が走行する走路に対応する信号機を対象信号機として抽出し、抽出した対象信号機のみを対象として、対象信号機ごとに検出領域を設定するものであってもよい。例えば、検出領域設定部140は、車両の位置、及び、車両が走行する道路の構造を示す地図情報に基づいて、車両が走行する走路に対応する信号機を対象信号機として抽出する。これにより、後述する判定部150、判定結果記憶部160、統合部165、出力決定部170、出力部180による処理により、対象信号機に係る結果列、又は、統合列に基づいて、対象信号機に基づく現示状態を出力することが可能となる。
The detection
また、検出領域設定部140は、対象信号機の中に「連動する信号機」があるかどうかを判定する。ここで「連動する信号機」とは、同じ停止線に対応する複数の信号機のことである。検出領域設定部140は、直接、地図情報から信号機に対応する停止線の位置を取得し、停止線の位置が同じであるか否かに基づいて、複数の信号機が「複数の信号機」であるか否かを判定してもよい。検出領域設定部140は、信号機の位置情報に基づいて、複数の信号機が「複数の信号機」であるか否かを判定してもよい。また、検出領域設定部140は、車載装置により検出した複数の信号機の位置情報に基づいて、複数の信号機が「複数の信号機」であるか否かを判定してもよい。
The detection
判定部150は、信号機ごとに設定された検出領域に対して画像処理を実行して、検出領域内の信号機を検出し、信号機の現示状態を判定する。判定部150は、例えば、テンプレートマッチングにより信号機を検出する。テンプレートマッチングは、標準の信号機の画像をテンプレートとして、それを1画素ずつずらしながら検出領域を走査し、例えば輝度の分布の相関を計算する。そして、相関が最も高い値となった場合にテンプレートがある画像上の位置に信号機があると検出する。
The
信号機が示す「色信号」には、「青信号」、「黄信号」、「赤信号」が含まれる。「色信号」が表す意味は、車両が従うべき交通法規等によって定まり、例えば、「青信号」は「進んでも良い」の意味を表し、「赤信号」は、「停止位置で止まれ」の意味を表し、「黄信号」は、「停止位置に近接しているため安全に停止することができない場合を除き、停止位置で止まれ」の意味を表す。このように、「青信号」は、車両の「発進許可」の意味を有し、「赤信号」及び「黄信号」は、車両の「発進許可」以外の意味を有している。 The "color signals" shown by traffic lights include "green light," "yellow light," and "red light." The meaning of a "color signal" is determined by traffic laws that vehicles must follow; for example, a "green light" means "you may proceed," a "red light" means "stop at the stop position," and a "yellow light" means "stop at the stop position, unless you are too close to the stop position to stop safely." In this way, a "green light" means "permission to start" for a vehicle, while a "red light" and a "yellow light" have meanings other than "permission to start" for a vehicle.
このような「青信号」、「黄信号」、「赤信号」の識別は、3つの「色信号」のうち、最も輝度レベルの高い「色信号」が点灯しているとして行うものであってもよい。 Such discrimination between "green light," "yellow light," and "red light" may be performed by assuming that the "color signal" with the highest brightness level among the three "color signals" is illuminated.
その他、信号機は「色信号」のみならず、信号機が設置された交差点において車両に許可される方向を示す「矢印信号」を示すものであってもよい。「右折信号」「直進信号」「左折信号」などが挙げられる。 In addition, traffic lights are not limited to "color signals," but may also display "arrow signals" that indicate the direction vehicles are permitted to go at the intersection where the traffic lights are installed. Examples include "right turn signals," "straight ahead signals," and "left turn signals."
「矢印信号」は、「右折信号」「直進信号」「左折信号」にとどまらず、信号機が設置された交差点の構造に応じて様々なバリエーションが考えられる。「矢印信号」が表す意味は、車両が従うべき交通法規等によって定まる。 "Arrow signals" are not limited to "right turn signals," "straight ahead signals," and "left turn signals," and there are many variations depending on the structure of the intersection where the traffic lights are installed. The meaning of an "arrow signal" is determined by the traffic laws that vehicles must follow.
判定部150は、検出領域に対して画像処理を実行して、信号機の現示状態として、信号機の「色信号」や「矢印信号」の点灯状態を判定する。
The
なお、判定部150での信号機検出の画像処理は、サポートベクターマシンやニューラルネットワークなどの機械学習を利用してもよい。信号機を検出する場合、予め大きさの異なる信号機のテンプレートが格納された学習データベースを備えておき、信号機までの距離に応じて参照する学習データベースを使い分けることで認識率の向上が図れる。
In addition, the image processing for traffic light detection in the
判定結果記憶部160は、判定部150によって判定された一又は複数の信号機の現示状態を、信号機ごとに記憶する。具体的には、撮像部71によって取得された時系列順の複数の画像に対して、信号機ごとに判定部150が順次判定を行い、得られた判定結果を判定結果記憶部160は判定部150から取得する。そして、判定結果記憶部160は、信号機ごとの時系列順の複数の判定結果を、信号機に対応付けられた結果列として記憶する。
The determination
図4を用いて、判定結果記憶部160によって記憶される結果列の構造について説明する。図4は、時系列順の複数の判定結果からなる、1つの信号機についての結果列の例を示す図である。以下では説明の簡略化のため、判定部150で得られる判定結果は、「青信号」、「黄信号」、「赤信号」、「不明」(現示状態が不明であるとする判定結果)の4種類であるとする。図4において、判定結果である「青信号」、「黄信号」、「赤信号」、「不明」は、それぞれ、白色で塗りつぶされた四角形、対角ストライプのパターンで塗りつぶされた四角形、市松模様のパターンで塗りつぶされた四角形、「?」印が付された四角形で示されている。以降の図面においても同じ表記を採用している。
The structure of the result string stored by the judgment
なお、判定部150による判定結果が「不明」となる理由としては、例えば、フリッカー現象(信号機ランプがLEDランプのように周期的に点滅しているランプである場合に、撮像部71による撮像タイミングに依存して撮像画像中の輝度が変化する現象)やローリングシャッター現象(撮像部71での撮像方式に起因して、撮像画像中の位置ごとに撮像タイミングがずれる現象)などにより、信号機がちらつく場合が挙げられる。
The reason why the
判定結果記憶部160によって記憶される、1つの信号機に対応付けられた1つの結果列を構成する判定結果の個数Nは、あらかじめ設定されている。
The number N of judgment results that constitute one result sequence associated with one traffic light and are stored by the judgment
1つの信号機に対応付けられた1つの結果列を構成するN個の判定結果には、それぞれ0からN-1までの添字が付与される。ここで添字0の判定結果は、判定部150によって得られた最新の判定結果であり、判定結果に付与された添字が増えるにしたがって、順に古い判定結果であることを意味する。
The N judgment results that make up one result string associated with one traffic light are each assigned a subscript ranging from 0 to N-1. Here, the judgment result with
判定結果記憶部160に新しい判定結果が1個入力されると、判定結果記憶部160は、判定結果に付与された添字を1だけ増加させ、判定結果記憶部160に入力された新しい判定結果に添字0を付与する。添字がNとなった判定結果は、消去される。
When a new judgment result is input to the judgment
このようにして、判定結果記憶部160には、0からN-1までの添字が付与された判定結果が、結果列として記憶される。
In this way, the judgment results with subscripts from 0 to N-1 are stored in the judgment
その他、判定結果記憶部160は、判定結果の補完機能を備えていてもよい。具体的には、判定結果記憶部160において記憶されている結果列を構成する判定結果の中に、「不明」が含まれている場合、当該「不明」の直前の判定結果と当該「不明」の直後の判定結果に基づいて、当該「不明」の判定結果を置き換えるものであってもよい。
In addition, the judgment
例えば、図4に示す結果列において、添字2の判定結果は「不明」となっている。この場合、判定結果記憶部160は、添字2の判定結果の直前にある添字3の判定結果と、添字2の判定結果の直後にある添字1の判定結果を参照する。添字1の判定結果と添字3の判定結果は両者ともに「黄信号」である。そのため、添字2の判定結果は「不明」ではなく、添字2の判定結果の直前にある添字3の判定結果である「黄信号」としてもよい。
For example, in the result string shown in FIG. 4, the judgment result for
このように、判定結果記憶部160は、結果列を構成する判定結果のうち、現示状態が不明であるとする判定結果を第1判定結果として抽出するものであってもよい。そして、判定結果記憶部160は、結果列を構成する判定結果のうち、現示状態が不明であるとする判定結果を除いた判定結果の中から、第1判定結果の直前の判定結果、及び、第1判定結果の直後の判定結果を、それぞれ、第2判定結果、及び、第3判定結果として抽出し、第2判定結果と第3判定結果が同一である場合に、結果列内の第1判定結果を第2判定結果で置き換えるものであってもよい。
In this way, the judgment
また、判定結果記憶部160は、結果列を構成する判定結果のうち、第1判定結果を除いた判定結果の中に第1判定結果の直前の判定結果が存在しない場合には、結果列内の第1判定結果を第1判定結果の直後の判定結果で置き換えるものであってもよい。これにより、結果列の先頭に現示状態が不明であるとする判定結果が存在する場合であっても、判定結果の補完を行うことができる。
In addition, the judgment
また、判定結果記憶部160は、結果列を構成する判定結果のうち、第1判定結果を除いた判定結果の中に第1判定結果の直後の判定結果が存在しない場合には、結果列内の第1判定結果を第1判定結果の直前の判定結果で置き換えるものであってもよい。これにより、結果列の末尾に現示状態が不明であるとする判定結果が存在する場合であっても、判定結果の補完を行うことができる。
In addition, the judgment
信号機のちらつきにより、判定部150によって信号機の現示状態を判定できない場合が生じる状況において、上述した判定結果記憶部160による判定結果の補完機能は、信号機の認識精度を向上させるのに有効である。
In situations where the
その他、後述する統合部165によって統合列が生成された後、判定結果記憶部160は、連動する信号機がある状態からない状態に変化したか否かを判定するものであってもよい。そして、連動する信号機がある状態からない状態に変化したと判定された場合に、判定結果記憶部160は、信号機に係る結果列を構成する判定結果を、統合列を構成する判定結果で置き換えるものであってもよい。
In addition, after the integrated sequence is generated by the
そして、判定結果記憶部160に新しい判定結果が1個入力されると、統合列を構成する判定結果で置き換えた後の結果列に対して、判定結果記憶部160は、判定結果に付与された添字を1だけ増加させ、判定結果記憶部160に入力された新しい判定結果に添字0を付与するものであってもよい。添字がNとなった判定結果は、消去される。
When one new judgment result is input to the judgment
これにより、画像上の信号機が複数から1つになり、統合列に基づく多数決判定から1つの結果列に基づく多数決判定に戻る場合において、非常に短いステップの間に、信号機の現示状態の判定結果が頻繁に変化してしまうことが抑制される。その結果、現示状態の判定結果が安定化し、現示状態の判定結果を車両制御の用に用いやすくなる。 This prevents the current signal status judgment result from changing too frequently in a very short step when the number of traffic lights on the image goes from multiple to one, and the majority decision based on the integrated column goes back to majority decision based on a single result column. As a result, the current signal status judgment result becomes more stable, making it easier to use the current signal status judgment result for vehicle control.
上述の説明では、判定結果記憶部160に記憶される判定結果はN個であるとして説明したが、これに限定されない。
In the above explanation, the judgment results stored in the judgment
例えば、判定結果記憶部160は、判定結果の個数の代わりに、判定結果の元となった画像の撮像時刻に基づいて、判定結果を記憶するか否かを決定してもよい。より具体的には、判定結果記憶部160は、撮像してから所定時間以上経過した撮像画像に対応する判定結果を消去するようにして、判定結果記憶部160に記憶される判定結果は、現在から所定時間過去までに撮像された画像に対応するもののみであるとしてもよい。
For example, the judgment
撮像部71によって画像を所定周期で撮像する際には、判定結果記憶部160に記憶される判定結果の個数を、個数Nに基づいて制限する場合と、所定時間に基づいて制限する場合とで結果は同じである。
When the
一方、撮像部71によって画像を所定周期ではない不特定のタイミングで撮像する際には、個数Nに基づいて制限する場合と、所定時間に基づいて制限する場合とで結果が変わりうる。しかしながら、記憶される判定結果の個数を、個数Nに基づいて制限する場合と、所定時間に基づいて制限する場合のいずれの場合であっても、判定結果記憶部160の構成として採用できる。
On the other hand, when the
出力決定部170は、信号機ごとの結果列に含まれる判定結果、又は、後述する統合列に含まれる判定結果に基づいて、出力値を決定する。
The
まず初めに、出力決定部170は、連動する信号機がない場合、例えば、対象となる信号機の数が1である場合に、信号機に係る結果列を構成する判定結果の中で最も多い判定結果に係る現示状態を、出力値として設定する。すなわち、結果列を構成する判定結果の中で多数決によって定まる判定結果に係る現示状態を、出力値として設定する。
First, when there is no linked traffic light, for example when there is only one target traffic light, the
なお、最も多い判定結果が1つに定まらない場合には、出力値として設定する際に優先して用いる判定結果に係る現示状態をあらかじめ設定してもよい。例えば、車両が確実に停止することを優先するために、「赤信号」、「黄信号」、「青信号」、「不明」の順番で優先して出力値に設定するものであってもよい。 In addition, if the most common judgment result is not determined to be one, the current status related to the judgment result to be used preferentially when setting the output value may be set in advance. For example, to give priority to ensuring that the vehicle stops, the output value may be set preferentially in the following order: "red light," "yellow light," "green light," and "unknown."
図5を参照して、画像上の信号機の数が1である場合の出力決定部170の動作を説明する。図5は、1つの信号機に係る結果列に基づく現示状態の出力の例を示す図である。図5では、1つの信号機に係る結果列が、ステップ「t」からステップ「t+13」まで遷移する様子が示されている。
The operation of the
ステップ「t」からステップ「t+4」までの間では、結果列を構成する判定結果のうち、「赤信号」に係る判定結果が最も多いため、出力決定部170は、出力値として「赤」を設定する。
Between step "t" and step "t+4", the majority of the judgment results in the result sequence are related to "red light", so the
ステップ「t+5」では、「赤信号」に係る判定結果と「青信号」に係る判定結果が同数となっており、最も多い判定結果が1つに定まらない。そこで、出力決定部170は、「青信号」よりも「赤信号」を優先させて、出力決定部170は、出力値として「赤」を設定する。
In step "t+5", there are the same number of judgment results related to "red light" and "green light", and it is not possible to determine which is the most common judgment result. Therefore, the
ステップ「t+6」からステップ「t+11」までの間では、結果列を構成する判定結果のうち、「青信号」に係る判定結果が最も多いため、出力決定部170は、出力値として「青」を設定する。
Between step "t+6" and step "t+11", the majority of the judgment results in the result sequence are related to "green light", so the
ステップ「t+12」からステップ「t+13」までの間では、結果列を構成する判定結果のうち、「黄信号」に係る判定結果が最も多いため、出力決定部170は、出力値として「黄」を設定する。
Between step "t+12" and step "t+13", the majority of the judgment results in the result sequence are related to "yellow light", so the
さらに、出力決定部170は、連動する信号機がある場合に、後述する統合列を構成する判定結果の中で最も多い判定結果に係る現示状態を、出力値として設定する。すなわち、統合列を構成する判定結果の中で多数決によって定まる判定結果に係る現示状態を、出力値として設定する。出力決定部170によって行われる、統合列に基づく出力値の決定方法は、上述した結果列に基づく出力値の決定方法と同様である。
Furthermore, when there are interlocking traffic signals, the
上述の出力決定部170の説明では、連動する信号機がない場合には、結果列に基づいて出力値を決定し、連動する信号機がある場合には、統合列に基づいて出力値を決定するものとして説明した。代わりに、連動する信号機がない場合には、統合部165が、結果列の判定結果を統合列に上書きするものであってもよい。これにより、出力決定部170は、信号機の数に関わらず常に、統合列に基づいて出力値を決定するものであってもよい。
In the above description of the
統合部165は、連動する信号機がある場合に、連動する信号機ごとの結果列を、連動する信号機にわたって統合して統合列を生成する。例えば、一の連動する信号機に係る結果列を構成する判定結果のうち、最新の判定結果を含む連続する所定個数以下の判定結果からなる部分列を、連動する信号機にわたって統合して統合列を生成する。
When there are interlocking traffic signals, the
なお、統合部165は、「部分列のサイズ算出処理」として、一の連動する信号機に係る結果列を構成する判定結果の数を連動する信号機の数で除算して得られる値以下の値を所定個数として設定するものであってもよい。
In addition, the
図6A、図6Bを参照して、統合列を生成する統合部165の動作を説明する。図6Aは、2つの結果列を統合する場合の例を示す図である。図6Bは、3つの結果列を統合する場合の例を示す図である。
The operation of the merging
図6Aでは、第1信号機に係る結果列、及び、第2信号機に係る結果列が、ステップ「t」からステップ「t+6」まで遷移する様子が示されている。図6Bでは、さらに、第3信号機に係る結果列が、ステップ「t」からステップ「t+6」まで遷移する様子が示されている。 FIG. 6A shows the transition of the result sequence related to the first traffic light and the result sequence related to the second traffic light from step "t" to step "t+6". FIG. 6B further shows the transition of the result sequence related to the third traffic light from step "t" to step "t+6".
図6Aでは、統合部165は、所定個数が5(結果列を構成する判定結果の数を、信号機の数である2で除算して得られる値)に設定される例が示されている。すなわち、図6Aでは、各ステップにおいて部分列は、添字0~4が付された5つの判定結果によって構成されている。
In FIG. 6A, the
図6Aの「統合列の遷移」に示されるように、統合部165は、第1信号機に係る結果列の一部分である部分列と、第2信号機に係る結果列の一部分である部分列とを、そのまま連結して、統合列を生成している。例えば、第1信号機の部分列の添字0~4の判定結果を、それぞれ統合列の添字0~4の判定結果としている。また、第2信号機の部分列の添字0~4の判定結果を、それぞれ統合列の添字5~9の判定結果としている。
As shown in "Transition of integrated sequence" in FIG. 6A, the
第1信号機に係る結果列に基づいて、出力決定部170が判定を行う場合、ステップ「t」からステップ「t+4」までの間の多数決結果は「青」となり、ステップ「t+5」以降の多数決結果は「黄」となる。第2信号機に係る結果列に基づく多数決結果も同様である。
When the
一方、統合列に基づいて、出力決定部170が判定を行う場合、ステップ「t」からステップ「t+2」までの間の多数決結果は「青」となり、ステップ「t+3」以降の多数決結果は「黄」となる。したがって、結果列に基づいて判定を行う場合と比較して、統合列に基づいて判定を行う場合には、2ステップ分だけ、信号機の現示状態が切り替わってから、現示状態が切り替わったと実際に判定されるまでの遅延が抑制されることになる。
On the other hand, if the
信号機の現示状態が切り替わってから、現示状態が切り替わったと実際に判定されるまでの遅延を抑制する効果は、統合列の生成に寄与する信号機の数が多くなるほど大きくなる。 The effect of reducing the delay between when the current signal status of a traffic light changes and when it is actually determined that the current signal status has changed becomes greater the more traffic lights that contribute to the generation of the integrated sequence.
図6Bでは、統合部165は、所定個数が3(結果列を構成する判定結果の数を、信号機の数である3で除算して得られる値を超えない最も大きな整数)に設定される例が示されている。すなわち、図6Bでは、各ステップにおいて部分列は、添字0~2が付された3つの判定結果によって構成されている。
In FIG. 6B, the
図6Bの「統合列の遷移」に示されるように、統合部165は、第1信号機に係る結果列の一部分である部分列と、第2信号機に係る結果列の一部分である部分列と、第3信号機に係る結果列の一部分である部分列とを、そのまま連結して、統合列を生成している。例えば、第1信号機の部分列の添字0~2の判定結果を、それぞれ統合列の添字0~2の判定結果としている。また、第2信号機の部分列の添字0~2の判定結果を、それぞれ統合列の添字3~5の判定結果としている。さらに、第3信号機の部分列の添字0~2の判定結果を、それぞれ統合列の添字6~8の判定結果としている。
As shown in "Transition of integrated sequence" in FIG. 6B, the
統合列に基づいて、出力決定部170が判定を行う場合、ステップ「t」からステップ「t+1」までの間の多数決結果は「青」となり、ステップ「t+2」以降の多数決結果は「黄」となる。したがって、結果列に基づいて判定を行う場合と比較して、統合列に基づいて判定を行う場合には、3ステップ分だけ、信号機の現示状態が切り替わってから、現示状態が切り替わったと実際に判定されるまでの遅延が抑制されることになる。
When the
上述の図6A、図6Bでは、結果列の一部である部分列を統合して統合列を生成する例を説明した。しかしながら、必ずしも部分列を設定して統合する必要はない。例えば、統合部165は、結果列をそのまま統合して統合列を生成してもよい。添字0~9が付された判定結果から構成される結果列を2つ統合する場合、統合列は添字0~19が付された判定結果から構成されることになる。
In the above-mentioned Figures 6A and 6B, an example was described in which a subsequence that is part of a result sequence is integrated to generate an integrated sequence. However, it is not necessary to set a subsequence and integrate it. For example, the
その他、統合部165は、連動する信号機に係る結果列を構成する判定結果の時系列順に基づいて、統合列を構成する判定結果を時系列順に並べて統合列を生成するものであってもよい。
In addition, the
図7を参照して、時系列順を考慮して統合列を生成する統合部165の動作を説明する。図7は、時系列順を考慮して2つの結果列を統合する場合の例を示す図である。
The operation of the integrating
図7における第1信号機に係る結果列及び部分列、第2信号機に係る結果列及び部分列の様子は、図6A、図6Bに示される様子と同じである。 The result sequence and subsequence for the first traffic light in Figure 7, and the result sequence and subsequence for the second traffic light in Figure 7 are the same as those shown in Figures 6A and 6B.
図7の「統合列の遷移」に示されるように、統合部165は、第1信号機に係る結果列の一部分である部分列と、第2信号機に係る結果列の一部分である部分列とを、時系列順を考慮して組み合わせて統合列を生成している。例えば、第1信号機の部分列の添字0,1,2,3,4の判定結果を、それぞれ統合列の添字0,2,4,6,8の判定結果としている。また、第2信号機の部分列の添字0,1,2,3,4の判定結果を、それぞれ統合列の添字1,3,5,7,9の判定結果としている。
As shown in "Transition of integrated sequence" in Figure 7, the
このように時系列順を考慮して統合列を生成することにより、多数決結果に加えて、時系列順の情報に基づいて、出力決定部170は出力値の決定を行うことができる。また、信号機に係る結果列を構成する判定結果を、統合列を構成する判定結果で置き換える場合に、置換後の結果列における判定結果において、時系列順が失われてしまう状況を解消することができる。
By generating the integrated sequence in this way while taking the chronological order into consideration, the
これにより、画像上の信号機が複数から1つになった場合において、統合列に基づく多数決判定から1つの結果列に基づく多数決判定に戻る場合において、非常に短いステップの間に、信号機の現示状態の判定結果が頻繁に変化してしまうことが抑制される。その結果、現示状態の判定結果が安定化し、現示状態の判定結果を車両制御の用に用いやすくなる。 This prevents the current signal status judgment result from changing too frequently in a very short step when the number of traffic lights on the image goes from multiple to one, and the process returns from majority judgment based on the integrated sequence to majority judgment based on a single result sequence. As a result, the current signal status judgment result becomes more stable, making it easier to use the current signal status judgment result for vehicle control.
例えば、図8は、連動する信号機が減少した場合の現示状態の出力の例を示す図である。図8では、第1信号機に係る結果列が、ステップ「t」からステップ「t+6」まで遷移する様子が示されている。また、第2信号機に係る結果列が、ステップ「t」からステップ「t+3」まで遷移する様子が示されている。ステップ「t+4」において、第2信号機が画像上から消えて、画像上には第1信号機のみが写りこんでいる状況であるとする。 For example, Figure 8 is a diagram showing an example of the current status output when the number of linked traffic lights is reduced. Figure 8 shows how the result sequence related to the first traffic light transitions from step "t" to step "t+6". Also, it shows how the result sequence related to the second traffic light transitions from step "t" to step "t+3". At step "t+4", the second traffic light disappears from the image, and only the first traffic light is visible on the image.
図8では、統合部165は、所定個数が5(結果列を構成する判定結果の数を、信号機の数である2で除算して得られる値)に設定される例が示されている。すなわち、図8では、各ステップにおいて部分列は、添字0~4が付された5つの判定結果によって構成されている。
In FIG. 8, the
ステップ「t+4」において、出力決定部170は、統合列に基づく多数決判定から、第1信号機の結果列に基づく多数決判定に戻る。
In step "t+4", the
ここで、第1信号機に係る結果列を構成する判定結果を、統合列を構成する判定結果で置き換えない場合、問題が生じる。つまり、ステップ「t+3」において統合列に基づく多数決判定が「黄」であるのに対し、ステップ「t+4」において、第1信号機の結果列に基づく多数決判定が「青」になっている。つまり、非常に短いステップの間に、信号機の現示状態の判定結果が、「青」→「黄」→「青」のように、頻繁に変化してしまう。 Here, a problem arises if the judgment results constituting the result sequence for the first traffic light are not replaced with the judgment results constituting the integrated sequence. That is, in step "t+3", the majority judgment based on the integrated sequence is "yellow", whereas in step "t+4", the majority judgment based on the result sequence for the first traffic light is "green". In other words, in a very short number of steps, the judgment result of the current state of the traffic light changes frequently, such as "green" -> "yellow" -> "green".
一方、第1信号機に係る結果列を構成する判定結果を、統合列を構成する判定結果で置き換えた場合、上記問題は生じない。図8の「置換後の第1信号機の結果列の推移」において、ステップ「t+3」の結果列の判定結果は、ステップ「t+3」の統合列の判定結果によって置き換えられている。例えば、ステップ「t+3」において、統合列の添字0~9の判定結果を、それぞれ第1信号機の結果列の添字0~9の判定結果としている。
On the other hand, if the judgment results constituting the result sequence for the first traffic light are replaced with the judgment results constituting the integrated sequence, the above problem does not occur. In the "Progression of the result sequence for the first traffic light after replacement" in Figure 8, the judgment results of the result sequence in step "t+3" are replaced with the judgment results of the integrated sequence in step "t+3". For example, in step "t+3", the judgment results of the
このように統合列を構成する判定結果で置き換えた結果、ステップ「t+4」において、第1信号機の結果列に基づく多数決判定を「黄」とすることができる。そのため、非常に短いステップの間に、信号機の現示状態の判定結果が頻繁に変化してしまうことを抑制できる。 As a result of replacing the judgment results that make up the integrated sequence in this way, in step "t+4", the majority decision based on the result sequence of the first traffic light can be "yellow". This makes it possible to prevent the judgment results of the current state of the traffic light from changing frequently in a very short step.
信号機の現示状態の判定結果が頻繁に変化してしまうことを抑制する効果は、図7で示したように、時系列順を考慮して統合列を生成する場合により大きくなる。 The effect of preventing frequent changes in the judgment results of the current signal state of a traffic light is greater when the integrated sequence is generated taking into account the chronological order, as shown in Figure 7.
その他、統合部165は、前回の統合列の生成の後に連動する信号機が増加したか否かを判定するものであってもよい。統合部165は、連動する信号機が増加したと判定された場合に、新たに追加された連動する信号機に係る結果列を構成する判定結果の数が所定閾値以上となるまで、新たに追加された連動する信号機以外の、連動する信号機ごとの結果列を統合して統合列を生成するものであってもよい。
In addition, the integrating
なお、統合部165は、一の信号機に係る結果列を構成する判定結果の数を信号機の数で除算して得られる値以下の値を所定閾値として設定するものであってもよい。
The
図9を参照して、信号機が増加した場合の統合部165の動作を説明する。図9は、連動する信号機が増加した場合の現示状態の出力の例を示す図である。
The operation of the
図9では、第1信号機に係る結果列、及び、第2信号機に係る結果列が、ステップ「t」からステップ「t+6」まで遷移する様子が示されている。ただし、ステップ「t」よりも前のステップでは画像上に写りこんでいなかった第2信号機が、ステップ「t」以降では画像上に写りこんでいる状況であるとする。第2信号機は、ステップ「t」で画像に写りこんだため、第2信号機に係る結果列を構成する判定結果のうち、添字0の判定結果を除く、添字1~9の判定結果は「不明」となっている。
FIG. 9 shows how the result sequence related to the first traffic light and the result sequence related to the second traffic light transition from step "t" to step "t+6". However, assume that the second traffic light, which was not captured in the image in steps prior to step "t", is captured in the image from step "t" onwards. As the second traffic light was captured in the image in step "t", the judgment results that make up the result sequence related to the second traffic light, with the
図9では、統合部165は、所定閾値が5(結果列を構成する判定結果の数を、信号機の数である2で除算して得られる値)に設定される例が示されている。
In FIG. 9, the
このような状況の場合、統合部165は、第2信号機に係る結果列を構成する判定結果の数が所定閾値以上となるまで、第2信号機に係る結果列を統合列の生成に用いない。図9に示す例では、ステップ「t+4」の時点で、第2信号機に係る結果列を構成する判定結果の数が5となり、所定閾値以上となっている。そこで、統合部165は、ステップ「t」からステップ「t+3」までの間は、第2信号機に係る結果列を統合列の生成に用いない。一方、統合部165は、ステップ「t+4」以降に、第2信号機に係る結果列を統合列の生成に用いる。
In such a situation, the integrating
このように、新たに追加された信号機についての結果列に、十分な数の判定結果が蓄積されるまで、統合部165は、新たに追加された信号機についての結果列を、統合列の生成に用いないものであってもよい。これにより、統合列に基づく多数決判定の精度が低下することが抑制される。
In this way, the
出力部180は、出力決定部170によって決定された出力値を、「信号機に基づく現示状態」として出力する。ここで、「信号機に基づく現示状態」とは、一又は複数の信号機の現示状態を統合することによって得られた結果であるといえる。
The
例えば、「信号機に基づく現示状態」は、出力部180から車両制御装置400に出力され、車両の制御に使用される。その他、「信号機に基づく現示状態」は、出力部180から図示しない通報装置に出力され、通報装置を介して乗員に通知されてもよい。
For example, the "current status based on the traffic light" is output from the
[信号機認識装置の処理手順]
次に、本実施形態に係る信号機認識装置による信号機認識の処理手順を、図2及び図3のフローチャートを参照して説明する。図2は、本実施形態に係る信号機認識装置の処理手順を示すフローチャートである。図3は、出力決定処理を示すフローチャートである。
[Processing procedure of traffic light recognition device]
Next, a process for traffic light recognition by the traffic light recognition device according to the present embodiment will be described with reference to the flowcharts of Figures 2 and 3. Figure 2 is a flowchart showing the process by the traffic light recognition device according to the present embodiment. Figure 3 is a flowchart showing an output determination process.
図2及び図3に示す信号機認識の処理は、撮像部71で画像を取得するたびに実行されるものであってもよいし、撮像部71で画像を取得した後、検出領域に対して画像処理を実行する周期ごとに実行されるものであってもよい。
The traffic light recognition process shown in Figures 2 and 3 may be executed each time an image is captured by the
図2では、ステップS101において、コントローラ100は、車両の進行方向にある一又は複数の信号機を撮像した画像を取得する。
In FIG. 2, in step S101, the
ステップS103において、コントローラ100は、車載センサ73等を介して、車両の位置を取得する。
In step S103, the
ステップS105において、コントローラ100は、地図情報取得部75等から取得した地図情報に基づいて、一又は複数の信号機の位置を取得する。
In step S105, the
ステップS107において、検出領域設定部140は、車両が走行する走路に対応する信号機を対象信号機として抽出する。なお、取得した画像に映りこむ全ての信号機を対象信号機とする場合、ステップS103、S105、S107の処理は省略されてもよい。
In step S107, the detection
ステップS109において、判定結果記憶部160は、連動する信号機がある状態からない状態に変化したか否かを判定する。
In step S109, the determination
連動する信号機がある状態からない状態に変化した場合(ステップS109でYESの場合)、ステップS111で、判定結果記憶部160は、信号機に係る結果列を統合列で置換する。すなわち、判定結果記憶部160は、信号機に係る結果列を構成する判定結果を、統合列を構成する判定結果で置き換える。
If the state changes from one in which there is an interlocking traffic light to one in which there is no interlocking traffic light (YES in step S109), in step S111, the determination
一方、連動する信号機がある状態からない状態に変化していない場合(ステップS109でNOの場合)、ステップS111をスキップして、ステップS113に進む。 On the other hand, if the state of the interlocking traffic light has not changed from a state where there is an interlocking traffic light to a state where there is no interlocking traffic light (NO in step S109), step S111 is skipped and the process proceeds to step S113.
ステップS113において、検出領域設定部140は、画像上に、信号機ごとに検出領域を設定する。
In step S113, the detection
ステップS115において、判定部150は、信号機ごとに設定された検出領域に対して画像処理を実行して検出領域内の信号機を検出し、信号機ごとに現示状態を判定する。
In step S115, the
ステップS117において、判定結果記憶部160は、信号機ごとに結果列を記憶する。
In step S117, the determination
ステップS119において、出力決定部170は、図3に示す出力決定処理を行う。
In step S119, the
図3では、ステップS121において、出力決定部170は、連動する信号機があるか否かを判定する。
In FIG. 3, in step S121, the
連動する信号機がある場合(ステップS121でYESの場合)、ステップS123において、出力決定部170は、「部分列のサイズ算出処理」を行う。
If there is an interlocking traffic light (YES in step S121), in step S123, the
ステップS125において、統合部165は、設定された部分列に基づいて統合列を生成する。
In step S125, the
ステップS127において、出力決定部170は、統合列に基づいて出力値を決定する。
In step S127, the
一方、連動する信号機がない場合(ステップS121でNOの場合)、ステップS131において、出力決定部170は、一の信号機の結果列に基づいて出力値を決定する。
On the other hand, if there is no linked traffic light (NO in step S121), in step S131, the
ステップS127、又は、ステップS131の処理の後、ステップS129において、出力決定部170は、出力値を出力部180に出力する。その後、出力部180は、出力決定部170によって決定された出力値を、「信号機に基づく現示状態」として出力する。
After the processing of step S127 or step S131, in step S129, the
[信号機認識装置の変形例]
図10、図11を用いて、信号機認識装置の変形例を説明する。
[Modification of the traffic light recognition device]
A modified example of the traffic light recognition device will be described with reference to Figs.
上記実施形態では、部分列のサイズ(部分列を構成する判定結果の数)は、複数の信号機の間で共通であるものとして説明した。また、部分列のサイズは、信号機の結果列を構成する判定結果の数に基づいて決定され、不変であるものとして説明した。 In the above embodiment, the size of the subsequence (the number of judgment results that make up the subsequence) has been described as being common among multiple traffic lights. Also, the size of the subsequence has been described as being determined based on the number of judgment results that make up the result sequence of the traffic light, and is invariant.
しかしながら、部分列のサイズを、信号機の結果列のうち最新の判定結果に基づいて変動させてもよい。例えば、統合部165は、連動する信号機がある場合に、一の連動する信号機に係る結果列を構成する判定結果のうち、最新の判定結果を含む連続する判定結果からなる部分列を、連動する信号機にわたって統合して統合列を生成するものであってもよい。その際、車両の発進許可を示している最新の判定結果を有する部分列を構成する判定結果の数よりも、車両の発進許可以外を示している最新の判定結果を有する部分列を構成する判定結果の数は小さいものであってもよい。
However, the size of the subsequence may be varied based on the latest judgment result in the traffic light result sequence. For example, when there are interlocking traffic lights, the
図10は、部分列のサイズ算出処理を示すフローチャートである。ステップS201において、統合部165は、未選択である信号機の結果列のうち、一の結果列を選択する。
FIG. 10 is a flowchart showing the process of calculating the size of a subsequence. In step S201, the
ステップS203において、統合部165は、選択した結果列を構成する判定結果のうち、最新の判定結果(添字0の判定結果)が「青」であるか否かを判定する。つまり、統合部165は、車両の発進許可を示しているか否かを判定する。
In step S203, the
最新の判定結果が「青」である場合(ステップS203でYESの場合)、ステップS205において、統合部165は、選択した結果列に係る部分列のサイズとして第1個数を設定する。
If the latest judgment result is "blue" (YES in step S203), in step S205, the
一方、最新の判定結果が「青」でない場合(ステップS203でNOの場合)、ステップS207において、統合部165は、選択した結果列に係る部分列のサイズとして第2個数を設定する。ここで、第2個数は第1個数よりも小さい。
On the other hand, if the latest judgment result is not "blue" (NO in step S203), in step S207, the
ステップS209において、統合部165は、未選択の結果列が存在するか否かを判定する。
In step S209, the
未選択の結果列が存在する場合(ステップS209でYESの場合)、ステップS201に戻る。未選択の結果列が存在しない場合(ステップS209でNOの場合)、部分列のサイズ算出処理を終了する。 If there is an unselected result string (YES in step S209), return to step S201. If there is no unselected result string (NO in step S209), end the substring size calculation process.
上述の例では、最新の判定結果が「青」である場合と、最新の判定結果が「青」でない場合とで、それぞれ、第1個数、第2個数を部分列のサイズを設定したが、これに限定されない。例えば、最新の判定結果が「青」、「黄」、「赤」である場合のそれぞれに対応して、第1個数、第2個数、第3個数を設定するものであってもよい。ここで、第1個数よりも第2個数は小さく、第2個数よりも第3個数は小さいものであってもよい。 In the above example, the size of the subsequence is set to the first number and the second number for the cases where the latest judgment result is "blue" and where the latest judgment result is not "blue", but this is not limiting. For example, the first number, second number, and third number may be set to correspond to the cases where the latest judgment result is "blue", "yellow", and "red", respectively. Here, the second number may be smaller than the first number, and the third number may be smaller than the second number.
図11は、部分列のサイズを変動させて2つの結果列を統合する場合の例を示す図である。図11では、第1信号機に係る結果列、及び、第2信号機に係る結果列が、ステップ「t」からステップ「t+6」まで遷移する様子が示されている。 FIG. 11 shows an example of integrating two result sequences by varying the size of the subsequences. FIG. 11 shows how the result sequence related to the first traffic light and the result sequence related to the second traffic light transition from step "t" to step "t+6."
第1信号機に係る結果列に係る部分列のサイズを参照すると、ステップ「t」では、最新の判定結果は「青」であるため、部分列のサイズが5に設定されている。一方、ステップ「t+1」からステップ「t+6」までの間では、最新の判定結果は「黄」であるため、部分列のサイズが3に設定されている。 Referring to the size of the subsequence related to the result sequence related to the first traffic light, at step "t", the most recent judgment result is "green", so the size of the subsequence is set to 5. On the other hand, between step "t+1" and step "t+6", the most recent judgment result is "yellow", so the size of the subsequence is set to 3.
第2信号機に係る結果列に係る部分列のサイズを参照すると、ステップ「t」、ステップ「t+1」では、最新の判定結果は「青」であるため、部分列のサイズが5に設定されている。一方、ステップ「t+2」からステップ「t+6」までの間では、最新の判定結果は「黄」であるため、部分列のサイズが3に設定されている。 Referring to the size of the subsequence related to the result sequence for the second traffic light, at step "t" and step "t+1", the most recent judgment result is "green", so the size of the subsequence is set to 5. On the other hand, from step "t+2" to step "t+6", the most recent judgment result is "yellow", so the size of the subsequence is set to 3.
その結果、統合列のサイズ(統合列を構成する判定結果の数)は、ステップ「t」において10、ステップ「t+1」において8、ステップ「t+2」からステップ「t+6」までの間において6となっている。 As a result, the size of the integrated column (the number of judgment results that make up the integrated column) is 10 at step "t", 8 at step "t+1", and 6 from step "t+2" to step "t+6".
第1信号機に係る結果列に基づいて、出力決定部170が判定を行う場合、ステップ「t」からステップ「t+4」までの間の多数決結果は「青」となり、ステップ「t+5」以降の多数決結果は「黄」となる。第2信号機に係る結果列に基づいて、出力決定部170が判定を行う場合、ステップ「t」からステップ「t+5」までの間の多数決結果は「青」となり、ステップ「t+6」以降の多数決結果は「黄」となる。
When the
これに対し、上述のように生成された統合列に基づいて、出力決定部170が判定を行う場合、ステップ「t」からステップ「t+2」までの間の多数決結果は「青」となり、ステップ「t+3」以降の多数決結果は「黄」となる。したがって、結果列に基づいて判定を行う場合と比較して、統合列に基づいて判定を行う場合には、3ステップ分だけ、信号機の現示状態が切り替わってから、現示状態が切り替わったと実際に判定されるまでの遅延が抑制されることになる。
In contrast, when the
このように、車両の発進許可を示している最新の判定結果を有する部分列を構成する判定結果の数よりも、車両の発進許可以外を示している最新の判定結果を有する部分列を構成する判定結果の数は小さく設定している。その結果、車両の発進許可以外の判定を行うまでの遅延をより効果的に抑制することができる。 In this way, the number of judgment results constituting the subsequence whose latest judgment result indicates something other than permission to start the vehicle is set to be smaller than the number of judgment results constituting the subsequence whose latest judgment result indicates permission to start the vehicle. As a result, the delay until a judgment other than permission to start the vehicle is made can be more effectively suppressed.
[実施形態の効果]
以上詳細に説明したように、本実施形態に係る信号機認識方法及び信号機認識装置は、車両の進行方向を撮像した複数枚の画像を対象として、画像ごとに信号機ごとの現示状態を判定し、時系列順の複数の判定結果からなる結果列を信号機ごとに取得する。複数の信号機の中に連動する信号機がある場合に、連動する信号機ごとの結果列を、連動する信号機にわたって統合して統合列を生成し、統合列を構成する判定結果の中で最も多い判定結果に係る現示状態を、信号機に基づく現示状態として出力する。
[Effects of the embodiment]
As described above in detail, the traffic light recognition method and traffic light recognition device according to this embodiment target a plurality of images captured in the direction of travel of the vehicle, determine the current status of each traffic light for each image, and obtain a result sequence consisting of a plurality of determination results in chronological order for each traffic light. If there are interlocking traffic lights among the plurality of traffic lights, the result sequences for each interlocking traffic light are integrated across the interlocking traffic lights to generate an integrated sequence, and the current status associated with the most frequent determination result among the determination results constituting the integrated sequence is output as the current status based on the traffic light.
これにより、信号機の現示状態が切り替わってから、現示状態が切り替わったと実際に判定されるまでの遅延を抑制することができる。特に、結果列を構成する判定結果に基づく多数決結果よりも、判定結果の切り替わりを早めることができる。また、複数の信号機の現示状態の判定結果に基づいて、より精度の高い現示状態の判定を行うことができる。 This makes it possible to reduce the delay between when the current status of a traffic light changes and when it is actually determined that the current status has changed. In particular, it is possible to speed up the change in the judgment result compared to a majority vote result based on the judgment results that make up the result string. Also, it is possible to make a more accurate judgment of the current status based on the judgment results of the current status of multiple traffic lights.
また、本実施形態に係る信号機認識方法及び信号機認識装置は、連動する信号機がある場合に、一の連動する信号機に係る結果列を構成する判定結果のうち、最新の判定結果を含む連続する所定個数以下の判定結果からなる部分列を、連動する信号機にわたって統合して統合列を生成するものであってもよい。これにより、複数の信号機の現示状態の判定結果に基づいて、より精度の高い現示状態の判定を行うことができる。 The traffic light recognition method and traffic light recognition device according to this embodiment may also generate an integrated sequence by integrating a subsequence of a predetermined number or less of consecutive judgment results, including the most recent judgment result, among the judgment results that make up the result sequence related to one linked traffic light, across the linked traffic lights when there are linked traffic lights. This allows for more accurate determination of the current indication status based on the judgment results of the current indication status of multiple traffic lights.
さらに、本実施形態に係る信号機認識方法及び信号機認識装置は、連動する信号機がある場合に、一の連動する信号機に係る結果列を構成する判定結果の数を連動する信号機の数で除算して得られる値以下の値を所定個数として設定するものであってもよい。これにより、統合列のサイズ(統合列を構成する判定結果の数)が、結果列のサイズ(結果列を構成する判定結果の数)以下となる。その結果、結果列を構成する判定結果に基づく多数決結果よりも、統合列を構成する判定結果に基づく多数決結果において判定結果の切り替わりを早めることができる。 Furthermore, the traffic light recognition method and traffic light recognition device according to this embodiment may, when there are interlocking traffic lights, set the predetermined number to a value equal to or less than the value obtained by dividing the number of determination results that make up the result sequence for one interlocking traffic light by the number of interlocking traffic lights. This makes the size of the integrated sequence (the number of determination results that make up the integrated sequence) equal to or less than the size of the result sequence (the number of determination results that make up the result sequence). As a result, it is possible to switch the determination result more quickly in the majority decision result based on the determination results that make up the integrated sequence than in the majority decision result based on the determination results that make up the result sequence.
また、本実施形態に係る信号機認識方法及び信号機認識装置は、連動する信号機がある場合に、一の連動する信号機に係る結果列を構成する判定結果のうち、最新の判定結果を含む連続する判定結果からなる部分列を、連動する信号機にわたって統合して統合列を生成するものであってもよい。そして、車両の発進許可を示している最新の判定結果を有する部分列を構成する判定結果の数よりも、車両の発進許可以外を示している最新の判定結果を有する部分列を構成する判定結果の数は小さいものであってもよい。 The traffic light recognition method and traffic light recognition device according to this embodiment may also generate an integrated sequence by integrating, across all of the linked traffic lights, a subsequence of consecutive judgment results, including the most recent judgment result, among the judgment results constituting the result sequence related to one linked traffic light when there are linked traffic lights. The number of judgment results constituting the subsequence whose most recent judgment result indicates something other than permission to start the vehicle may be smaller than the number of judgment results constituting the subsequence whose most recent judgment result indicates permission to start the vehicle.
これにより、車両の発進許可以外の判定を行うまでの遅延をより効果的に抑制することができる。車両の発進許可の判定を行う場合の遅延よりも、車両の発進許可以外の判定を行う場合の遅延の方が小さくなる。その結果、車両の走行制御に影響を及ぼす可能性の高い信号機の現示状態をより効果的に判定することができる。 This makes it possible to more effectively reduce delays in making decisions other than those permitting the vehicle to start. The delay in making decisions other than those permitting the vehicle to start is smaller than the delay in making decisions permitting the vehicle to start. As a result, it is possible to more effectively determine the current status of traffic lights that are likely to affect vehicle driving control.
さらに、本実施形態に係る信号機認識方法及び信号機認識装置は、前回の統合列の生成の後に連動する信号機が増加したか否かを判定するものであってもよい。連動する信号機が増加したと判定された場合に、新たに追加された連動する信号機に係る結果列を構成する判定結果の数が所定閾値以上となるまで、新たに追加された連動する信号機以外の、連動する信号機ごとの結果列を統合して統合列を生成するものであってもよい。これにより、新たに追加された信号機に係る結果列を構成する判定結果の数が少ないことに起因して、統合列に基づく多数決判定の精度が低下することが抑制される。 Furthermore, the traffic light recognition method and traffic light recognition device according to this embodiment may determine whether the number of interlocking traffic lights has increased since the previous generation of the integrated sequence. If it is determined that the number of interlocking traffic lights has increased, the integrated sequence may be generated by integrating the result sequences of the interlocking traffic lights other than the newly added interlocking traffic light until the number of determination results constituting the result sequence for the newly added interlocking traffic light reaches or exceeds a predetermined threshold. This prevents a decrease in the accuracy of the majority decision based on the integrated sequence due to a small number of determination results constituting the result sequence for the newly added traffic light.
また、本実施形態に係る信号機認識方法及び信号機認識装置は、連動する信号機がある場合に、連動する信号機に係る結果列を構成する判定結果の時系列順に基づいて、統合列を構成する判定結果を時系列順に並べて統合列を生成するものであってもよい。これにより、多数決結果に加えて、時系列順の情報に基づいて、より精度の高い現示状態の判定を行うことができる。また、信号機に係る結果列を構成する判定結果を、統合列を構成する判定結果で置き換える場合に、置換後の結果列における判定結果において、時系列順が失われてしまう状況を解消することができる。 The traffic light recognition method and traffic light recognition device according to this embodiment may also generate an integrated sequence by arranging the determination results that make up the integrated sequence in chronological order based on the chronological order of the determination results that make up the result sequence related to the linked traffic lights when there are linked traffic lights. This allows for more accurate determination of the current state based on chronological order information in addition to the majority vote result. Also, when replacing the determination results that make up the result sequence related to a traffic light with the determination results that make up the integrated sequence, it is possible to eliminate a situation in which the chronological order is lost in the determination results in the result sequence after replacement.
さらに、本実施形態に係る信号機認識方法及び信号機認識装置は、統合列の生成の後に連動する信号機がある状態からない状態に変化したか否かを判定するものであってもよい。連動する信号機がある状態からない状態に変化したと判定された場合に、信号機に係る結果列を構成する判定結果を、統合列を構成する判定結果で置き換えるものであってもよい。これにより、非常に短いステップの間に、信号機の現示状態の判定結果が頻繁に変化してしまうことが抑制され、その結果、現示状態の判定結果を車両制御の用に用いやすくなる。 Furthermore, the traffic light recognition method and traffic light recognition device according to this embodiment may determine whether an interlocking traffic light has changed from a state in which it is present to a state in which it is not present after the generation of the integrated sequence. If it is determined that an interlocking traffic light has changed from a state in which it is present to a state in which it is not present, the determination result constituting the result sequence related to the traffic light may be replaced with the determination result constituting the integrated sequence. This prevents the determination result of the current status of the traffic light from changing frequently in a very short step, and as a result, it becomes easier to use the determination result of the current status for vehicle control.
また、本実施形態に係る信号機認識方法及び信号機認識装置は、信号機の数が1である場合に、信号機に係る結果列を構成する判定結果の中で最も多い判定結果に係る現示状態を、信号機に基づく現示状態として出力するものであってもよい。これにより、信号機の数が1である場合には、結果列を構成する判定結果を用いた判定を行うことができる。 The traffic light recognition method and traffic light recognition device according to this embodiment may also be configured to output, when there is one traffic light, the current indication state associated with the most frequent judgment result among the judgment results constituting the result string related to the traffic lights as the current indication state based on the traffic light, when there is one traffic light. This allows a judgment to be made using the judgment results constituting the result string when there is one traffic light.
さらに、本実施形態に係る信号機認識方法及び信号機認識装置において、連動する信号機は、車両が走行する走路における同じ停止線に対応する複数の信号機であってもよい。これにより、車両が走行する上で従うべき信号機のみを対象として現示状態の判定を行うことができる。その結果、信号機の現示状態の誤認識を抑制できる。 Furthermore, in the traffic light recognition method and traffic light recognition device according to this embodiment, the interlocking traffic lights may be multiple traffic lights that correspond to the same stop line on the road on which the vehicle is traveling. This makes it possible to determine the current status of only the traffic lights that the vehicle should obey when traveling. As a result, erroneous recognition of the current status of traffic lights can be suppressed.
また、本実施形態に係る信号機認識方法及び信号機認識装置は、結果列を構成する判定結果のうち、現示状態が不明であるとする判定結果を第1判定結果として抽出するものであってもよい。結果列を構成する判定結果のうち、第1判定結果を除いた判定結果の中から、第1判定結果の直前の判定結果、及び、第1判定結果の直後の判定結果を、それぞれ、第2判定結果、及び、第3判定結果として抽出するものであってもよい。第2判定結果と第3判定結果が同一である場合に、結果列内の第1判定結果を第2判定結果で置き換えるものであってもよい。 The traffic light recognition method and traffic light recognition device according to this embodiment may extract, from among the judgment results constituting the result string, a judgment result that indicates that the current display state is unknown as the first judgment result. From among the judgment results constituting the result string, excluding the first judgment result, the judgment result immediately before the first judgment result and the judgment result immediately after the first judgment result may be extracted as the second judgment result and the third judgment result, respectively. If the second judgment result and the third judgment result are identical, the first judgment result in the result string may be replaced with the second judgment result.
これにより、信号機のちらつきにより、信号機の現示状態を判定できない場合が生じる状況においても、信号機の現示状態が実際に変化してから出力値が変化するまでの時間遅れを抑制でき、その結果、車両の空走距離を低減することができる。 This makes it possible to reduce the time delay between when the actual signal state of the traffic light changes and when the output value changes, even in situations where the current signal state cannot be determined due to flickering of the traffic light, thereby reducing the free-running distance of the vehicle.
例えば、図12に示すように、結果列を構成する判定結果に基づいて信号機の現示状態を判定する場合に、判定の遅延を抑制することができる。図12では、補間前の結果列によれば、出力決定部170は、ステップ「t+7」で初めて出力値として「黄」を設定できる。一方、補間後の結果列によれば、出力決定部170は、ステップ「t+6」で初めて出力値として「黄」を設定できる。このように、判定の遅延を抑制することができる。
For example, as shown in FIG. 12, when determining the current state of a traffic light based on the determination results that constitute a result sequence, delays in determination can be suppressed. In FIG. 12, according to the result sequence before interpolation, the
さらに、本実施形態に係る信号機認識方法及び信号機認識装置は、結果列を構成する判定結果のうち、第1判定結果を除いた判定結果の中に第1判定結果の直前の判定結果が存在しない場合には、結果列内の第1判定結果を第1判定結果の直後の判定結果で置き換えるものであってもよい。これにより、結果列の先頭に現示状態が不明であるとする判定結果が存在する場合であっても、判定結果の補完を行うことができる。 Furthermore, the traffic light recognition method and traffic light recognition device according to this embodiment may replace the first judgment result in the result string with the judgment result immediately following the first judgment result if there is no judgment result immediately preceding the first judgment result among the judgment results constituting the result string, excluding the first judgment result. This makes it possible to complement the judgment results even if there is a judgment result at the beginning of the result string indicating that the current status is unknown.
また、本実施形態に係る信号機認識方法及び信号機認識装置は、結果列を構成する判定結果のうち、第1判定結果を除いた判定結果の中に第1判定結果の直後の判定結果が存在しない場合には、結果列内の第1判定結果を第1判定結果の直前の判定結果で置き換えるものであってもよい。これにより、結果列の先頭に現示状態が不明であるとする判定結果が存在する場合であっても、判定結果の補完を行うことができる。 The traffic light recognition method and traffic light recognition device according to this embodiment may also replace the first judgment result in the result string with the judgment result immediately preceding the first judgment result if there is no judgment result immediately following the first judgment result among the judgment results constituting the result string, excluding the first judgment result. This makes it possible to complement the judgment results even if there is a judgment result at the beginning of the result string that indicates that the current status is unknown.
上述の実施形態で示した各機能は、1又は複数の処理回路によって実装されうる。処理回路には、プログラムされたプロセッサや、電気回路などが含まれ、さらには、特定用途向けの集積回路(ASIC)のような装置や、記載された機能を実行するよう配置された回路構成要素なども含まれる。 Each of the functions described in the above embodiments may be implemented by one or more processing circuits. Processing circuits include programmed processors, electrical circuits, and even devices such as application specific integrated circuits (ASICs), and circuit components arranged to perform the described functions.
以上、実施形態に沿って本発明の内容を説明したが、本発明はこれらの記載に限定されるものではなく、種々の変形及び改良が可能であることは、当業者には自明である。この開示の一部をなす論述および図面は本発明を限定するものであると理解すべきではない。この開示から当業者には様々な代替実施形態、実施例および運用技術が明らかとなろう。 The contents of the present invention have been described above in accordance with the embodiments, but the present invention is not limited to these descriptions, and it will be obvious to those skilled in the art that various modifications and improvements are possible. The descriptions and drawings that form part of this disclosure should not be understood as limiting the present invention. Various alternative embodiments, examples, and operating techniques will be apparent to those skilled in the art from this disclosure.
本発明はここでは記載していない様々な実施形態等を含むことは勿論である。したがって、本発明の技術的範囲は上記の説明から妥当な特許請求の範囲に係る発明特定事項によってのみ定められるものである。 The present invention naturally includes various embodiments not described here. Therefore, the technical scope of the present invention is determined only by the invention-specific matters related to the scope of the claims that are appropriate from the above explanation.
71 撮像部
73 車載センサ
75 地図情報取得部
100 コントローラ
140 検出領域設定部
150 判定部
160 判定結果記憶部
165 統合部
170 出力決定部
180 出力部
400 車両制御装置
71
Claims (13)
前記コントローラは、
前記撮像部を用いて、前記車両の進行方向にある一又は複数の信号機を、複数回撮像して複数の画像を取得し、
前記画像ごとに前記信号機ごとの現示状態を判定して得られた、時系列順の複数の判定結果からなる結果列を前記信号機ごとに取得し、
前記複数の信号機の中に連動する信号機がある場合に、
前記連動する信号機ごとの前記結果列を、前記連動する信号機にわたって統合して統合列を生成し、
前記統合列を構成する判定結果の中で最も多い判定結果に係る現示状態を、前記信号機に基づく現示状態として出力すること
を特徴とする信号機認識方法。 A traffic light recognition method for controlling a controller connected to an imaging unit mounted on a vehicle, comprising:
The controller:
Using the imaging unit, one or more traffic lights in a traveling direction of the vehicle are imaged a plurality of times to obtain a plurality of images;
A result sequence is obtained for each of the traffic lights, the result sequence being made up of a plurality of judgment results in chronological order, the judgment results being obtained by judging the current state of each of the traffic lights for each of the images;
When there is an interlocking signal among the plurality of signals,
aggregating the result sequence for each of the interlocking traffic lights across the interlocking traffic lights to generate an aggregate sequence;
A traffic light recognition method, comprising the steps of: outputting a current indication state associated with the most frequent judgment result among the judgment results constituting the integrated sequence as a current indication state based on the traffic light.
一の前記連動する信号機に係る前記結果列を構成する判定結果のうち、最新の判定結果を含む連続する所定個数以下の判定結果からなる部分列を、前記連動する信号機にわたって統合して前記統合列を生成すること
を特徴とする、請求項1に記載の信号機認識方法。 The controller, when the interlocking traffic light is present,
The traffic light recognition method according to claim 1, characterized in that, among the judgment results constituting the result sequence relating to one of the interlocking traffic lights, a subsequence consisting of a predetermined number or less of consecutive judgment results including the most recent judgment result is integrated across the interlocking traffic lights to generate the integrated sequence.
一の前記連動する信号機に係る前記結果列を構成する判定結果の数を前記連動する信号機の数で除算して得られる値以下の値を前記所定個数として設定すること
を特徴とする、請求項2に記載の信号機認識方法。 The controller, when the interlocking traffic light is present,
The traffic light recognition method according to claim 2, characterized in that the predetermined number is set to a value equal to or less than a value obtained by dividing the number of judgment results constituting the result sequence relating to one of the interlocking traffic lights by the number of the interlocking traffic lights.
一の前記連動する信号機に係る前記結果列を構成する判定結果のうち、最新の判定結果を含む連続する判定結果からなる部分列を、前記連動する信号機にわたって統合して前記統合列を生成し、
前記車両の発進許可を示している最新の判定結果を有する前記部分列を構成する判定結果の数よりも、前記車両の発進許可以外を示している最新の判定結果を有する前記部分列を構成する判定結果の数は小さいこと
を特徴とする、請求項1~3のいずれか一項に記載の信号機認識方法。 The controller, when the interlocking traffic light is present,
generating the integrated sequence by integrating a subsequence of successive determination results including a latest determination result among the determination results constituting the result sequence related to one of the interlocking traffic lights across the interlocking traffic lights;
A traffic light recognition method as described in any one of claims 1 to 3, characterized in that the number of judgment results constituting the subsequence having the latest judgment result indicating something other than permission to start the vehicle is smaller than the number of judgment results constituting the subsequence having the latest judgment result indicating permission to start the vehicle.
前回の前記統合列の生成の後に前記連動する信号機が増加したか否かを判定し、
前記連動する信号機が増加したと判定された場合に、新たに追加された前記連動する信号機に係る前記結果列を構成する判定結果の数が所定閾値以上となるまで、新たに追加された前記連動する信号機以外の、前記連動する信号機ごとの前記結果列を統合して前記統合列を生成すること
を特徴とする、請求項1~4のいずれか一項に記載の信号機認識方法。 The controller:
determining whether the number of interlocking traffic lights has increased since the last generation of the integrated sequence;
The traffic light recognition method according to any one of claims 1 to 4, characterized in that when it is determined that the number of interlocking traffic lights has increased, the integrated sequence is generated by integrating the result sequences for each of the interlocking traffic lights other than the newly added interlocking traffic light until the number of determination results that constitute the result sequence related to the newly added interlocking traffic light becomes equal to or greater than a predetermined threshold.
前記連動する信号機に係る前記結果列を構成する判定結果の時系列順に基づいて、前記統合列を構成する判定結果を時系列順に並べて前記統合列を生成すること
を特徴とする、請求項1~5のいずれか一項に記載の信号機認識方法。 The controller, when the interlocking traffic light is present,
A traffic light recognition method as described in any one of claims 1 to 5, characterized in that the integrated sequence is generated by arranging the judgment results that constitute the integrated sequence in chronological order based on the chronological order of the judgment results that constitute the result sequence related to the linked traffic lights.
前記統合列の生成の後に前記連動する信号機がある状態からない状態に変化したか否かを判定し、
前記連動する信号機がある状態からない状態に変化したと判定された場合に、前記信号機に係る前記結果列を構成する判定結果を、前記統合列を構成する判定結果で置き換えること
を特徴とする、請求項1~6のいずれか一項に記載の信号機認識方法。 The controller:
determining whether the interlocking traffic signal has changed from a present state to a non-present state after the generation of the integrated sequence;
A traffic light recognition method as described in any one of claims 1 to 6, characterized in that when it is determined that the linked traffic light has changed from a present state to a absent state, the judgment result constituting the result sequence relating to the traffic light is replaced with the judgment result constituting the integrated sequence.
前記信号機に係る前記結果列を構成する判定結果の中で最も多い判定結果に係る現示状態を、前記信号機に基づく現示状態として出力すること
を特徴とする、請求項1~7のいずれか一項に記載の信号機認識方法。 When the number of the traffic lights is 1, the controller
A traffic light recognition method as described in any one of claims 1 to 7, characterized in that the current indication status associated with the most frequent judgment result among the judgment results constituting the result sequence related to the traffic light is output as the current indication status based on the traffic light.
を特徴とする、請求項1~8のいずれか一項に記載の信号機認識方法。 The traffic light recognition method according to any one of claims 1 to 8, characterized in that the interlocking traffic lights are a plurality of traffic lights corresponding to the same stop line on the road on which the vehicle is traveling.
前記結果列を構成する判定結果のうち、前記現示状態が不明であるとする判定結果を第1判定結果として抽出し、
前記結果列を構成する判定結果のうち、前記第1判定結果を除いた判定結果の中から、前記第1判定結果の直前の判定結果、及び、前記第1判定結果の直後の判定結果を、それぞれ、第2判定結果、及び、第3判定結果として抽出し、
前記第2判定結果と前記第3判定結果が同一である場合に、前記結果列内の前記第1判定結果を前記第2判定結果で置き換えること
を特徴とする、請求項1~9のいずれか一項に記載の信号機認識方法。 The controller:
extracting a determination result indicating that the current state is unknown as a first determination result from among the determination results constituting the result sequence;
extracting, from among the judgment results constituting the result sequence, a judgment result immediately before the first judgment result and a judgment result immediately after the first judgment result as a second judgment result and a third judgment result, respectively;
A traffic light recognition method according to any one of claims 1 to 9, characterized in that when the second judgment result and the third judgment result are identical, the first judgment result in the result sequence is replaced with the second judgment result.
前記結果列を構成する判定結果のうち、前記現示状態が不明であるとする判定結果を第1判定結果として抽出し、
前記結果列を構成する判定結果のうち、前記第1判定結果を除いた判定結果の中に前記第1判定結果の直前の判定結果が存在しない場合には、前記結果列内の前記第1判定結果を前記第1判定結果の直後の判定結果で置き換えること
を特徴とする、請求項1~9のいずれか一項に記載の信号機認識方法。 The controller:
extracting a determination result indicating that the current state is unknown as a first determination result from among the determination results constituting the result sequence;
A traffic light recognition method according to any one of claims 1 to 9, characterized in that if there is no judgment result immediately preceding the first judgment result among the judgment results constituting the result sequence, excluding the first judgment result, the first judgment result in the result sequence is replaced with the judgment result immediately following the first judgment result.
前記結果列を構成する判定結果のうち、前記現示状態が不明であるとする判定結果を第1判定結果として抽出し、
前記結果列を構成する判定結果のうち、前記第1判定結果を除いた判定結果の中に前記第1判定結果の直後の判定結果が存在しない場合には、前記結果列内の前記第1判定結果を前記第1判定結果の直前の判定結果で置き換えること
を特徴とする、請求項1~9のいずれか一項に記載の信号機認識方法。 The controller:
extracting a determination result indicating that the current state is unknown as a first determination result from among the determination results constituting the result sequence;
A traffic light recognition method according to any one of claims 1 to 9, characterized in that if there is no judgment result immediately following the first judgment result among the judgment results constituting the result sequence, excluding the first judgment result, the first judgment result in the result sequence is replaced with the judgment result immediately preceding the first judgment result.
前記コントローラは、
前記撮像部を用いて、前記車両の進行方向にある一又は複数の信号機を、複数回撮像して複数の画像を取得し、
前記画像ごとに前記信号機ごとの現示状態を判定して得られた、時系列順の複数の判定結果からなる結果列を前記信号機ごとに取得し、
前記複数の信号機の中に連動する信号機がある場合に、
前記連動する信号機ごとの前記結果列を、前記連動する信号機にわたって統合して統合列を生成し、
前記統合列を構成する判定結果の中で最も多い判定結果に係る現示状態を、前記信号機に基づく現示状態として出力すること
を特徴とする信号機認識装置。 A traffic light recognition device including an imaging unit and a controller mounted on a vehicle,
The controller:
Using the imaging unit, one or more traffic lights in a traveling direction of the vehicle are imaged a plurality of times to obtain a plurality of images;
A result sequence is obtained for each of the traffic lights, the result sequence being made up of a plurality of judgment results in chronological order, the judgment results being obtained by judging the current state of each of the traffic lights for each of the images;
When there is an interlocking signal among the plurality of signals,
aggregating the result sequence for each of the interlocking traffic lights across the interlocking traffic lights to generate an aggregate sequence;
A traffic light recognition device characterized by outputting the current indication status associated with the most frequent judgment result among the judgment results constituting the integrated sequence as the current indication status based on the traffic light.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| PCT/JP2023/002897 WO2024161469A1 (en) | 2023-01-30 | 2023-01-30 | Traffic signal recognition method and traffic signal recognition device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| PCT/JP2023/002897 WO2024161469A1 (en) | 2023-01-30 | 2023-01-30 | Traffic signal recognition method and traffic signal recognition device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| WO2024161469A1 true WO2024161469A1 (en) | 2024-08-08 |
Family
ID=92146185
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| PCT/JP2023/002897 Ceased WO2024161469A1 (en) | 2023-01-30 | 2023-01-30 | Traffic signal recognition method and traffic signal recognition device |
Country Status (1)
| Country | Link |
|---|---|
| WO (1) | WO2024161469A1 (en) |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20170024622A1 (en) * | 2015-07-24 | 2017-01-26 | Honda Motor Co., Ltd. | Surrounding environment recognition device |
| JP2020067703A (en) * | 2018-10-22 | 2020-04-30 | 日産自動車株式会社 | Traffic light recognition method and traffic light recognition device |
| US20220398853A1 (en) * | 2019-11-12 | 2022-12-15 | Nissan Motor Co., Ltd. | Traffic Signal Recognition Method and Traffic Signal Recognition Device |
-
2023
- 2023-01-30 WO PCT/JP2023/002897 patent/WO2024161469A1/en not_active Ceased
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20170024622A1 (en) * | 2015-07-24 | 2017-01-26 | Honda Motor Co., Ltd. | Surrounding environment recognition device |
| JP2020067703A (en) * | 2018-10-22 | 2020-04-30 | 日産自動車株式会社 | Traffic light recognition method and traffic light recognition device |
| US20220398853A1 (en) * | 2019-11-12 | 2022-12-15 | Nissan Motor Co., Ltd. | Traffic Signal Recognition Method and Traffic Signal Recognition Device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| JP7255707B2 (en) | Traffic light recognition method and traffic light recognition device | |
| JP7388971B2 (en) | Vehicle control device, vehicle control method, and vehicle control computer program | |
| US10339805B2 (en) | Traffic light recognition device and traffic light recognition method | |
| US8848980B2 (en) | Front vehicle detecting method and front vehicle detecting apparatus | |
| JP7255706B2 (en) | Traffic light recognition method and traffic light recognition device | |
| JP7381388B2 (en) | Signal lamp status identification device, signal lamp status identification method, computer program for signal lamp status identification, and control device | |
| KR20110113185A (en) | Method and apparatus for determining effective lane marking | |
| JP2005092857A (en) | Image processing system and vehicle control system | |
| CN104210421A (en) | Ambient environment determination apparatus | |
| JP2020067698A (en) | Marking line detection device and marking line detection method | |
| JP2018163438A (en) | Information processing device and information processing system | |
| CN103917411A (en) | Method and device for grouping lighting units | |
| EP4213112B1 (en) | Method and device for detecting a color of a traffic light in an environment of a vehicle | |
| WO2024161469A1 (en) | Traffic signal recognition method and traffic signal recognition device | |
| WO2024161521A1 (en) | Traffic signal recognition method and traffic signal recognition device | |
| CN114746923B (en) | Signal recognition method and signal recognition device | |
| JP2024030951A (en) | Vehicle control device, vehicle control method, and vehicle control computer program | |
| JP2024125880A (en) | Dark place determination device, dark place determination method, and dark place determination program | |
| JP2019219816A (en) | Image processing system and lighting fixture for vehicle | |
| JP2024011893A (en) | Lane determination device, lane determination direction, and computer program for lane determination | |
| US20080310679A1 (en) | Vehicle and Lane Mark Recognizer | |
| RU2779798C1 (en) | Traffic light recognition method and traffic light recognition device | |
| JP7805201B2 (en) | Object recognition method and object recognition device | |
| RU2779773C1 (en) | Traffic light recognition method and traffic light recognition device | |
| US20250225875A1 (en) | Notification system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| 121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 23919614 Country of ref document: EP Kind code of ref document: A1 |
|
| NENP | Non-entry into the national phase |
Ref country code: DE |