JP2018018215A - Object feature point detector - Google Patents

Abstract

The present invention relates to an object feature point detection device, and an object thereof is to correctly detect a point on a target object as a feature point even when a target object to be detected overlaps another object. SOLUTION: The travel route information of a vehicle 10 is detected. Recognize lanes on the track based on track information. Objects existing in the vicinity of the vehicle 10 (the preceding vehicle 32, the roadside wall 84, etc.) are detected. Based on the detection result, a point on the contour of the target object (the preceding vehicle 32) is detected. The target object (the preceding vehicle 32) is fitted to the recognized lane, and the point on the contour is connected to another object (the wall 84, the surrounding vehicles), and the detection condition is detected on one side of the target object. It is determined whether or not. When the joining condition is satisfied on one side (for example, the left side), a point belonging to the other side (for example, the right side) of the target object is set as the characteristic point 106. [Selection diagram] Fig. 13

Description

  The present invention relates to an object feature point detection apparatus, and more particularly to an object feature point detection apparatus suitable as an apparatus mounted on a vehicle in order to detect a feature point of a target object existing in the vicinity.

  Patent Document 1 discloses a vehicle object detection device using a laser radar. This apparatus includes a laser radar having a detection area in front of the vehicle. According to the laser radar, the shape of an object existing in the detection area can be detected. The conventional apparatus selects a detection point of an object according to the following rule when an object enters the course of the host vehicle from the lateral direction.

1. When only the tip of the object is in the detection area: Select the detection point on the tip side.
2. When only the rear end of the object is in the detection area: The detection point on the rear end side is selected.
3. When all of the objects are within the detection area: The center detection point is selected from the plurality of detection points.

  According to the above rule, a point that most accurately represents the velocity of the object under each condition can be selected as a detection point. For this reason, according to the above-described conventional apparatus, the moving speed of the object can be accurately estimated from when the object starts to enter the detection area of the laser radar until the object escapes from the detection area.

  In addition to the above Patent Document 1, the applicant recognizes Patent Document 2 below as a document related to the present invention.

JP 2010-249690 A Japanese Patent Laid-Open No. 05-196736

  Incidentally, the laser radar can detect various other objects such as a white line on the road and a preceding vehicle on the adjacent lane, in addition to the target object entering the course of the host vehicle. When the target object to be detected is close to another object, the contours of both may overlap and may be detected by the laser radar in a state where both are combined.

  For this reason, according to the apparatus of Patent Document 1, the position of the tip of the object may be detected erroneously under close conditions, such as when the tip of the target object overlaps a white line on the road. Similarly, according to this apparatus, when the rear end of the object overlaps with a white line or the like, a situation in which the rear end position is erroneously detected may occur. If the leading end position or the trailing end position is erroneously detected, the central detection point is not necessarily correct. As described above, in the apparatus described in Patent Document 1, there may occur a situation in which a desired detection point cannot be obtained correctly under close conditions such as a target object overlapping another object.

  The present invention has been made to solve the above-described problems. Even when a target object to be detected is close to another object and a point on the correct object cannot be detected, the point on the target object is not detected. An object of the present invention is to provide an object feature point detection device capable of correctly detecting a feature point.

In order to achieve the above object, the present invention is an object feature point detection apparatus,
A process for detecting the running path information of the own vehicle;
Lane recognition processing for recognizing a lane on the road based on the road information,
An object detection process for detecting objects existing around the host vehicle;
A target object detection process for detecting a point on the contour of the target object based on the result of the object detection process;
Whether or not a connection condition is established on one side of the target object by fitting the target object to the lane and detecting a point on the contour combined with another object different from the target object A join determination process for determining
A feature point selection process in which a point on the contour belonging to the other side of the target object is a feature point of the target object when the combining condition is satisfied on one side;
It is characterized by performing.

  Further, according to a preferred aspect of the present invention, when the lane in which the target object is traveling is the lane at the end of the runway, the combination determination process is performed when the combination condition is satisfied on the end side. Includes processing to determine.

According to another preferred embodiment of the present invention,
The target object detection process includes a process of detecting a position of the target object based on a result of the object detection process,
The object feature point detection device further executes another object detection processing for detecting a point on the contour of the other object and a position of the other object based on the result of the object detection processing,
In the combination determination process, when the other object exists on one side of the target object and the distance between them is equal to or less than a determination threshold, the combination condition is satisfied on the side where the other object exists. Including the process of determining that the

According to another preferred embodiment of the present invention,
The target object detection process detects points on the contour on both one side and the other side of the target object for each sampling period,
The object feature point detection device includes:
A process of calculating an estimated position of the point based on the position of the point on the contour determined one cycle before, the moving speed of the target object, and the sampling period for each sampling period;
A process for calculating the difference between the point on the contour on the one side and the estimated position of the point and the difference between the point on the contour on the other side and the estimated position of the point for each sampling period; ,
Performing the process of determining whether or not the combination condition is satisfied on both sides of the target object,
The feature point selection process includes a process in which a point on the contour having the smaller difference is used as a feature point of the target object when the combination condition is satisfied on both sides of the target object.

  According to the present invention, it is possible to estimate a situation where a target object is placed by applying the target object to a lane recognized based on the road information. Based on the situation, it can be determined whether or not the coupling condition is established on one side of the target object. Furthermore, when the joining condition is established on one side, a point belonging to the other side of the target object is set as a feature point. This feature point is a point on the target object that is not combined with another object. Therefore, according to the present invention, a point on the contour of the target object can be correctly detected as a feature point even in a situation where the target object overlaps with another object.

It is a figure which shows the hardware constitutions of the object feature point detection apparatus of Embodiment 1 of this invention. It is a figure for demonstrating the subject which Embodiment 1 of this invention solves. FIG. 3 is a diagram showing an example of a situation in front of the vehicle shown in FIG. 2 and an example of a detection result corresponding to the situation. It is a block block diagram of the object feature point detection apparatus of Embodiment 1 of this invention. It is a figure for demonstrating the 1st example of the process which the lane recognition part shown in FIG. 4 performs. It is a figure for demonstrating the 2nd example of the process which the lane recognition part shown in FIG. 4 performs. It is a figure for demonstrating the 3rd example of the process which the lane recognition part shown in FIG. 4 performs. It is a figure for demonstrating the 4th example of the process which the lane recognition part shown in FIG. 4 performs. It is a figure for demonstrating the 1st example of the process which the combination determination part shown in FIG. 4 performs. It is a figure for demonstrating the 2nd example of the process which the combination determination part shown in FIG. 4 performs. It is a figure for demonstrating the 3rd example of the process which the combination determination part shown in FIG. 4 performs. It is a figure for demonstrating the 1st example of the process which the feature point selection part shown in FIG. 4 performs. It is a figure for demonstrating the 2nd example of the process which the feature point selection part shown in FIG. 4 performs. It is a figure for demonstrating the 3rd example of the process which the feature point selection part shown in FIG. 4 performs. It is a figure for demonstrating the 4th example of the process which the feature point selection part shown in FIG. 4 performs.

[Hardware Configuration of Embodiment 1]
FIG. 1 is a diagram showing a hardware configuration of the object feature point detection apparatus according to the first embodiment of the present invention. As shown in FIG. 1, a stereo camera 12 is mounted on the vehicle 10. The stereo camera 12 can take a stereo image of the front of the vehicle 10 with a predetermined viewing angle. The stereo camera 12 can be replaced with a monocular camera.

  A plurality of LIDAR (Laser Imaging Detection and Ranging) units 14 are mounted on the vehicle 10 so as to surround the vehicle. According to the LIDAR unit 14, it is possible to detect the contour of an object existing in each detection region and the distance to the object.

  A plurality of millimeter wave radar units 16 are mounted on the vehicle 10 so as to surround the vehicle. Millimeter wave radar is less susceptible to raindrops and gases than imaging by a camera or LIDAR. The millimeter wave radar unit 16 can detect the distance to an object existing in each detection region and the relative speed between the object and the vehicle 10.

  Hereinafter, the stereo camera 12, the LIDAR unit 14 and the millimeter wave radar unit 16 described above are collectively referred to as “object sensors”. The vehicle 10 may be equipped with a sonar sensor as an “object sensor”. According to the sonar sensor, an object existing around the vehicle 10 can be detected by using ultrasonic waves.

  The vehicle 10 further includes a GPS (Global Positioning System) unit 18. According to the GPS unit 18, the current position / orientation of the vehicle 10 can be detected using GPS.

  The vehicle 10 is equipped with an ECU (Electronic Control Unit) 20. The above-mentioned “object sensor” and the detection signals of the GPS unit 18 are supplied to the ECU 20. Based on these signals, the ECU 20 can detect various objects existing around the vehicle 10, specifically, road-side components such as white lines and guardrails, and preceding vehicles.

[Basic Functions of Embodiment 1]
FIG. 2 is a diagram for explaining a problem associated with the basic operation of the apparatus according to the present embodiment. Each of the four figures shown in FIG. 2 shows the result of object detection in the vehicle 10. In each figure, the code | symbol 22 shows the runway of the vehicle 10. FIG. On the runway 22, white lines 24 and 26 are drawn at the left and right ends, respectively. Further, a broken line white line 28 that divides the lane is drawn between the white lines 24 and 26. A wall 30 is drawn on the right side of the white line 26 on the right side.

  The leftmost figure in FIG. 2 shows the result of object detection at time t-1. In the vehicle 10, the detection result by the LIDAR unit 14 is first acquired as a premise for obtaining this result. According to the detection result of the LIDAR unit 14, the contour and position of an object existing around the vehicle 10 can be detected. In addition to the LIDAR unit 14, the vehicle 10 detects objects by the millimeter wave radar unit 16 and the stereo camera 12, and integrates them to reduce false detection / non-detection of the objects and estimate the position / velocity. Improve accuracy. A rectangle indicated by reference numeral 34 in the leftmost figure indicates the “tracking result” of the preceding vehicle 32 obtained as a result of the integration.

  In order to grasp the state (position and speed) of the preceding vehicle 32, it is necessary to capture a point on the preceding vehicle 32 where no change in position occurs except for movement. Therefore, the vehicle 10 recognizes one point on the contour as a “feature point” when the contour of the preceding vehicle 32 is detected by the LIDAR unit 14. In the present embodiment, in principle, a point closest to the vehicle 10 among the obtained points on the contour (hereinafter, referred to as “nearest neighbor point”) is recognized as a feature point. The left end graphic of FIG. 2 shows a state where a point corresponding to the right rear end corner of the preceding vehicle 32 is recognized as the feature point 36 according to this rule.

  2 shows the detection result 38 and the nearest point 40 of the LIDAR unit 14 at time t. In addition to the preceding vehicle 32, the LIDAR unit 14 detects the white lines 24, 26, 28 and the wall 30 as objects. In a situation where the preceding vehicle 32 is close to the wall 30 and the preceding vehicle 32 and the wall 30 overlap when viewed from the vehicle 10, the measurement points of both may be combined. In this case, the nearest point 40 of the detection result 38 is not a right rear end angle of the preceding vehicle 32 but a point on the wall 30.

  2 shows the tracking result 34 generated based on the detection result 38 at the time t. The tracking result 34 is generated by reflecting information (distance, speed, vehicle width, etc.) obtained by the stereo camera 12 or the millimeter wave radar unit 16 on the nearest point 40. At this stage, the vehicle 10 misidentifies the position of the preceding vehicle 32.

  The rightmost figure in FIG. 2 shows a state in which the state of the preceding vehicle 32 is estimated at time t + 1 based on the misdetected tracking result 34. As shown in this figure, if the tracking result 34 is erroneously recognized, the vehicle 10 will thereafter erroneously recognize the state of the preceding vehicle 32.

  FIG. 3 shows a situation where the coupling shown in FIG. 2 may occur and an example of the tracking result. Here, the characteristic points of the truck that is the preceding vehicle are detected in combination with the guardrail.

[Features of Embodiment 1]
FIG. 4 is a block diagram for explaining the contents of processing executed by the ECU 20 in the present embodiment in order to prevent the above-described erroneous recognition. The ECU 20 includes a processor, a memory, and various interfaces, and implements the functions of various blocks shown in FIG. 4 by performing processing according to a program stored in the memory.

(Lane recognition part)
The configuration shown in FIG. 4 includes a lane recognition unit 50. The lane recognition unit 50 is realized by the ECU 20 executing a lane recognition process for recognizing the situation of lanes provided around the vehicle 10.

  FIG. 5 is a diagram for explaining lane recognition processing executed by the lane recognition unit 50 in the present embodiment. As shown in FIG. 5, the lane recognition unit 50 is provided with runway information 52. The road information 52 includes information on road components provided from the object sensor, position / orientation information of the vehicle 10 provided from the GPS unit 18, and map information stored in the ECU 20 in advance. To do.

Specifically, the lane recognition unit 50 is realized by the ECU 20 performing the following processing based on the traveling road information 52.
(1) The current position of the vehicle 10 on the map is specified by applying the information on the road component detected by the object sensor and the position / orientation information from the GPS to the map information.
(2) The lane boundary line 54 around the current position of the vehicle 10 is estimated from the lane information included in the map information.

  The lane recognition process described above is an example of a process for realizing the lane recognition unit 50. Hereinafter, another example of the lane recognition process will be described with reference to FIGS.

FIG. 6 is a diagram for explaining a second example of lane recognition processing that can be used in the present embodiment. The example illustrated in FIG. 6 is realized by the ECU 20 performing the following process based on the travel path information 52.
(1) The white line position around the vehicle is detected by at least one of the stereo camera 12 and the LIDAR unit 14;
(2) When the detection distance is short with respect to the object, extrapolate to the necessary distance from the estimated shape and estimate the lane boundary line.

FIG. 7 is a diagram for explaining a third example of lane recognition processing that can be used in the present embodiment. The example illustrated in FIG. 7 is realized by the ECU 20 performing the following process based on the travel path information 52.
(1) Specify the current position of the vehicle 10 as in FIG.
(2) A grid map 58 is set in the peripheral area including the vehicle 10,
(3) A stationary object grid 60 is set in a grid in which the object is observed many times as a stationary object at the same position based on the measurement result of the object sensor.
(4) Read the lane width 62 at the current position from the map information,
(5) The lane boundary line 64 is estimated based on the arrangement of the stationary object grid 60 and the lane width 62.

FIG. 8 is a diagram for explaining a fourth example of lane recognition processing that can be used in the present embodiment. The example illustrated in FIG. 8 is realized by the ECU 20 performing the following processing based on the travel path information 52.
(1) Specify the current position of the vehicle 10 as in FIG.
(2) Detecting road structures 66 such as walls and guardrails based on the measurement results of the object sensor,
(3) A road boundary 68 is calculated by performing curve matching on the road structure 66;
(4) Read the lane width 70 at the current position from the map information,
(5) The lane boundary line 72 is estimated based on the road boundary 68 and the lane width 70.

  Note that the lane recognition processes described in the above paragraphs [0031] to [0036] may be executed in parallel or sequentially. At that time, robustness may be improved by weighted integration of the results or switching according to the detection state.

(Lane placement determination unit)
The processing result of the lane recognition unit 50 described above is provided to the lane arrangement determination unit 74 as shown in FIG. The lane arrangement determination unit 74 is realized by the ECU 20 executing a lane arrangement determination process for determining a traveling lane for each of the “target object” and the “other object”. Here, the “target object” means an object that is detected by the ECU 20 in the processing cycle, and the “other object” is obtained by removing “target object” from all objects detected by the object sensor. It shall mean an object.

  As shown in FIG. 4, the lane arrangement determination unit 74 is provided with a target object detection result 76 and another object detection result 78. The target object detection result 76 includes information such as the position and shape of the “target object”. Further, the other object detection result 78 includes information such as the position and shape of the “other object”.

  In the lane arrangement determination unit 74, a lane arrangement determination process based on the target object detection result 76 and a lane arrangement determination process based on the other object detection result 78 are executed. Since both processes are substantially the same, here, the process based on the target object detection result 76 will be described as a representative example thereof.

In the lane arrangement determination process based on the target object detection result 76, the ECU 20 sequentially executes the following processes.
(1) Reading the position of the target object,
(2) Reading the information of the center line of each lane,
(3) calculating the distance between the representative position (for example, the center position) of the target object and the center line of each lane;
(4) The lane with the shortest distance is determined as the traveling lane of the target object.

  The lane arrangement determination process described above is an example of a process for realizing the lane arrangement determination unit 74. In addition, “determined by whether or not it is inside the right and left lane boundary line” “calculates a probability value from the distance to the outside of the boundary line and determines from the probability value” “object trajectory and lane shape Is determined based on the degree of coincidence ”. Hereinafter, another example of the lane arrangement determination process that can be used in the present embodiment will be described.

The second example of the lane arrangement determination process is realized by the ECU 20 sequentially executing the following processes.
(1) Reading the position of the target object,
(2) Reading information on the left and right borders of each lane,
(3) calculating the distance between the representative position (for example, the center position) of the target object and the left and right boundary lines of each lane;
(4) A lane in which the distance is recognized on both the right side of the left boundary line and the left side of the right boundary line is determined as the traveling lane of the target object.

The third example of the lane arrangement determination process is realized by the ECU 20 sequentially executing the following processes.
(1) Reading the position of the target object,
(2) Read out a map to which the driving probability of the target object is assigned for each lane,
(3) Determine the travel lane of the target object by fitting the position of the target object to the map.

The fourth example of the lane arrangement determination process is realized by the ECU 20 sequentially executing the following processes.
(1) Read the travel locus of the target object,
(2) Collate each lane with the shape of the above-mentioned traveling locus (for example, curvature) and calculate the residual,
(3) As a result of the above collation, the traveling lane of the target object having the smallest residual is determined.

  Note that the lane arrangement determination processes described in the above paragraphs [0040] to [0044] may be executed in parallel or sequentially. At that time, robustness may be improved by weighted integration of the results or switching according to the detection state.

(Join determination part)
As shown in FIG. 4, the processing result of the lane arrangement determination unit 74 is provided to the combination determination unit 80. The coupling determination unit 80 is realized by the ECU 20 executing a coupling determination process described later. In the combination determination process, it is determined whether or not a condition for detecting the target object combined with another object (referred to as “join condition”) exists on either the left or right side of the target object or both. .

Hereinafter, the details of the combination determination process will be described with reference to FIGS. 9 to 11.
FIG. 9 shows a first example in which the join determination process determines whether the join condition is satisfied. In FIG. 9, the vehicle 10 recognizes the preceding vehicle 32 as the target object 82, and recognizes the wall 84 on the left side of the road as another object. A broken-line rectangle indicated by reference numeral 86 in the figure is a detection result of the LIDAR unit 14 in the current processing cycle. The detection result 86 is a result of erroneous detection that the left side of the preceding vehicle 32 is combined with the wall 84.

  The erroneous detection shown in FIG. 9 occurs when the target object 82 is traveling in the leftmost lane of the traveling road. For this reason, when the combination determination unit 80 can determine that the target object 82 is traveling in the leftmost lane of the runway based on the processing result of the lane arrangement determination unit 74, “the combination condition is on the left side of the target object 82. It is judged to exist. For the same reason, when it is determined that the target object 82 is traveling in the rightmost lane of the track, it is determined that “the coupling condition exists on the right side of the target object 82”.

  FIG. 10 shows a second example in which the combination determination process determines whether the combination condition is satisfied. In FIG. 10, the vehicle 10 recognizes the preceding vehicle 32 as the target object 82. Further, the vehicle 10 recognizes the surrounding vehicles in the left end lane and the right end lane as other objects 88 and 90. A rectangle indicated by reference numeral 92 in the drawing indicates the tracking result of the other object 88 in the leftmost lane. A rectangle 94 indicated by a broken line is a detection result of the LIDAR unit 14 in the current processing cycle. The detection result 94 is a result of erroneous detection of the target object 82 combined with the other object 90.

  The erroneous detection shown in FIG. 10 occurs when the tracking result 92 is recognized close to the left lane of the target object 82. For this reason, when it is determined that the tracking result is recognized in the left lane of the target object 82 at a distance equal to or smaller than the threshold based on the processing result of the lane arrangement determination unit 74, the combination determination unit 80 It is determined that a join condition exists on the left side of “. For the same reason, when it is determined that the tracking result is recognized in the right lane of the target object 82 at a distance equal to or less than the threshold, it is determined that “the coupling condition exists on the right side of the target object 82”.

  FIG. 11 shows a third example in which the combination determination process determines whether the combination condition is satisfied. In FIG. 11, the vehicle 10 recognizes the preceding vehicle 32 as the target object 82 and recognizes the two preceding two-wheeled vehicles as the other objects 96 and 98. Here, the two other objects 96 and 98 are traveling on the same lane as the target object 82. A rectangle denoted by reference numeral 100 in the drawing indicates a tracking result of the preceding other object 96. A rectangle 102 indicated by a broken line is a detection result of the LIDAR unit 14 in the current processing cycle. The detection result 102 is a result of erroneous detection of the target object 82 combined with the other object 96.

  The erroneous detection shown in FIG. 11 occurs when the tracking result 100 is recognized close to the left side of the target object 82 in the travel lane. For this reason, when it is determined based on the processing result of the lane arrangement determination unit 74 that the tracking result has been recognized at a distance equal to or less than the threshold on the left side of the traveling lane of the target object 82, the combination determination unit 80 It is determined that a coupling condition exists on the left side of the object 82 ”. For the same reason, when it is determined that the tracking result is recognized on the right side in the travel lane of the target object 82 at a distance equal to or less than the threshold, it is determined that “the coupling condition exists on the right side of the target object 82”. The

The above determination function is realized by the ECU 20 performing the following processing based on the lane arrangement of the target object and other objects.
(1) Read the traveling lane of the target object,
(2) If the driving lane is at the left end of the road, it is judged that “There is a coupling condition on the left side”.
(3) If the driving lane is at the right end of the track, it is determined that “There is a coupling condition on the right side”.
(4) Read the position of another object in the left lane of the target object,
(5) If the distance from the other object on the left lane is equal to or less than the threshold, it is determined that “the left side has a coupling condition”.
(6) Read the position of another object in the right lane of the target object,
(7) If the distance from the other object on the right lane is equal to or less than the threshold, it is determined that “the right side has a coupling condition”.
(8) Read out the position of another object that exists in the same lane as the traveling lane of the target object,
(9) If the distance between the other object in the same lane and the left side of the target object is less than the threshold, it is determined that “the left side has a coupling condition”.
(10) If the distance between the other object in the same lane and the right side of the target object is equal to or smaller than the threshold, it is determined that “the right side has a coupling condition”.
However, the process (2) may determine “there is a coupling condition” only when there is a wall on the left side of the runway. Similarly, the process (3) may determine “There is a coupling condition” only when there is a wall on the right side of the runway.

(Feature point selection part)
As illustrated in FIG. 4, the determination result of the combination determination unit 80 is provided to the feature point selection unit 104. The feature point selection unit 104 is realized by the ECU 20 executing a feature point selection process. In the feature point selection process, a process is performed in which a point on the target object that is not coupled to another object is used as a feature point based on the coupling condition of the target object.

Hereinafter, details of the feature point selection process will be described with reference to FIGS.
Specifically, in the feature point selection process, first, the following four cases are classified as to whether the joining condition is satisfied.
1. 1. There is no coupling condition on either side of the target object. There is a binding condition only on the left side of the target object,
3. There is a binding condition only on the right side of the target object,
4). There are coupling conditions on both the left and right sides of the target object.

  FIG. 12 is a diagram for explaining a process executed when the result 1 is obtained. When there is no connection condition on either the left or right side, the closest point of the point on the contour of the preceding vehicle 32 is determined as a feature point according to the principle. In the case shown in FIG. 12, the right rear end angle of the preceding vehicle 32 is selected as the feature point 106.

  FIG. 13 is a diagram for explaining the processing executed under the result 2 described above. When it is determined that there is a combination condition only on the left side of the preceding vehicle 32, a point on the opposite side of the detection result 108 by the LIDAR unit 14 on the side having the combination condition, that is, a point on the right side is set as a feature point. FIG. 13 shows a state in which the right rear end angle of the preceding vehicle 32 is selected as the feature point 106 in accordance with the above rules.

  FIG. 14 is a diagram for explaining the processing executed under the result 3 above. When it is determined that there is a coupling condition only on the right side of the preceding vehicle 32, the left point of the detection result 108 by the LIDAR unit 14 is set as a feature point. FIG. 14 shows a state in which the left rear end angle of the preceding vehicle 32 is selected as the feature point 106 in accordance with the rule.

  FIG. 15 is a diagram for explaining processing that is executed in the case of 4 above, that is, when there is a coupling condition on both sides of the preceding vehicle 32. FIG. 15 shows a state in which the right rear end corner of the preceding vehicle 32 is detected in combination with the wall 110 on the right side of the road. However, under the situation shown in FIG. 15, it must also be assumed that the left rear end angle of the preceding vehicle 32 is detected in combination with the left wall 112. Therefore, under such circumstances, first, the position of the left and right rear end corners of the current target object 82 is predicted based on the tracking result of the target object 82 detected in the previous processing cycle. Next, the difference between the left and right rear end angles in the detection result 108 obtained in the current processing cycle and the corresponding predicted position is calculated. Then, the rear end angle on the side where the difference from the predicted position is small is used as the feature point.

  In the situation shown in FIG. 15, the position of the right rear end corner of the detection result 108 is away from the original position of the right rear end corner of the target object 82 due to the coupling with the wall 110. The position of the left rear end corner of the detection result 108 correctly represents the position of the left rear end corner of the target object 82. In this case, the difference between the predicted value and the detection result 108 is small on the left side and large on the right side. Therefore, according to the above rules, the left rear end angle of the target object 82 is selected as the feature point 106 as shown in FIG. As described above, according to the above processing, even when there is a combination condition on both the left and right sides of the target object 82, a point on the contour of the target object 82 can be correctly selected as a feature point with high probability.

The feature point selection function described above is realized by the ECU 20 performing the following processing based on the coupling condition regarding the target object 82.
(1) Execute case classification related to join conditions,
(2) If there is no join condition on either side, select the nearest point as a feature point.
(3) When there is a join condition only on the left side, the right point of the detection result by the LIDAR unit 14 is selected as a feature point.
(4) If there is a join condition only on the right side, the left point of the detection result by the LIDAR unit 14 is selected as a feature point;
(5) If there are joint conditions on both the left and right sides,
(5-1) Read the tracking result at the previous processing cycle,
(5-2) Read the moving speed of the target object 82,
(5-3) Estimating the current position of the target vehicle from the read tracking result and moving speed, and the processing cycle period of the ECU 20;
(5-4) calculating the difference between the left rear end corner position of the target object 82 indicated by the detection result 108 of the LIDAR unit 14 in the current processing cycle and the estimated position of the point;
(5-5) calculating the difference between the right rear end corner position of the target object 82 indicated by the detection result 108 of the LIDAR unit 14 in the current processing cycle and the estimated position of the point;
(5-6) The difference calculated in the above (5-4) is compared with the difference calculated in the above (5-5), and the rear end corner position on the side where the difference is small is selected as the feature point.

  As shown in FIG. 4, the position of the feature point selected by the feature point selection unit 104 is determined as the position of the feature point in the current processing cycle. As described above, according to the object feature point detection device of the present embodiment, the contour of the target object can be detected even under a situation where the target object that is detected by the vehicle 10 appears to overlap another object other than the target object. The upper point can be correctly understood as a feature point. For this reason, if this device is used, the target object can be continuously captured accurately.

  In the description of the embodiments described in the above paragraphs [0022] to [0063], LIDAR is taken as an example, but the present invention is applicable to any sensor that can detect a “feature point” on the contour. Is possible. That is, even with a stereo camera or a millimeter wave radar, the same effect can be obtained in the same embodiment as long as the shape of the object can be estimated.

DESCRIPTION OF SYMBOLS 10 Vehicle 12 Stereo camera 14 LIDAR unit 16 Millimeter wave radar 18 GPS unit 20 ECU
22 Running track 24, 26, 54 White line 30, 84, 110, 112 Wall 32 Leading vehicle 34, 92, 100 Tracking result 36, 40, 106 Feature point 38, 86, 94, 102, 108 Detection result 50 Lane recognition unit 74 Lane Arrangement determining unit 80 Coupling determining unit 82 Target objects 88, 90, 96, 98 Other objects

Claims (1)

A process for detecting the running path information of the own vehicle;
Lane recognition processing for recognizing a lane on the road based on the road information,
An object detection process for detecting objects existing around the host vehicle;
A target object detection process for detecting a point on the contour of the target object based on the result of the object detection process;
Whether or not a connection condition is established on one side of the target object by fitting the target object to the lane and detecting a point on the contour combined with another object different from the target object A join determination process for determining
A feature point selection process in which a point on the contour belonging to the other side of the target object is a feature point of the target object when the combining condition is satisfied on one side;
An object feature point detection apparatus characterized by executing

Images