WO2002052523A1 - Method and apparatus for monitoring vehicle - Google Patents

Abstract

A dynamic image of a passing vehicle is picked up, in a specified frame cycle, by means of a CCD camera disposed in a movable housing installed on a roadside of a roadway and processed to determine the diameter of a wheel and the overall length of the vehicle and then the type of vehicle is determined from the determined diameter of the wheel and the overall length of the vehicle and recorded. Each frame of the dynamic image is processed. When a vehicle is detected from a vector generated through movement of a large number of feature points in the image, correlation between the shape of each vehicle type prepared in advance and the standard template of the tire of the type corresponding to the size is calculated about the tire image of the vehicle, a template having the smallest difference is found, and the type of the tire is determined.

Description

 Description Vehicle monitoring method and device

 The present invention relates to a method and apparatus for monitoring a vehicle that determines the type of a vehicle running on a road by performing image processing on an image captured by a camera installed on the side of a road, and particularly relates to a tire from a moving image of the vehicle. The present invention relates to a method and an apparatus for monitoring a vehicle for determining a vehicle type through a type. Background art

 Obtaining statistical data on vehicles traveling on the road is important for the operation and management of roads, traffic management, and so on. Work is being conducted to continuously count and investigate what vehicles have passed when. In this case, manual surveys are used to identify the vehicle types (large buses and trucks, medium-sized freight vehicles, and small vehicles) with a certain degree of accuracy due to fatigue caused by long hours of work and difficulty in distinguishing at night. It is difficult to do this, and it is desired to automate this.

 2. Description of the Related Art Conventionally, as a device for determining the type of a traveling vehicle, there is, for example, a vehicle type traffic volume measurement device employing a loop'sonic method. This device uses a loop coil and an ultrasonic head together as a vehicle sensing element. That is, the height of the vehicle is measured by the ultrasonic head, the length of the vehicle is measured by using the loop coil and the ultrasonic head together, and arithmetic processing is performed to classify passing vehicles. .

However, conventional devices that use the loop sonic method require that the loop antenna be embedded in the road, so that it cannot be used anywhere, for example, to measure the number of vehicles passing a day. . Therefore, at present, measurements for discriminating vehicles passing through an arbitrary location are performed manually, but it is difficult to perform counting with stable accuracy for 24 hours. There is a problem that it cannot be combined. Disclosure of the invention

 According to the present invention, there is provided a vehicle monitoring method and apparatus for collecting vehicle type data that is easily installed at an arbitrary location and that passes by unmanned.

 The present invention relates to a method for monitoring a vehicle traveling on a road, wherein a camera provided in a movable and installable casing installed on the side of a roadway is imaged at a predetermined frame cycle by a camera and a camera is used. It is characterized by measuring the diameter of the wheels and the total length of the vehicle by performing image processing on the acquired video, and discriminating and recording the type of vehicle from the measured diameters of the wheels and the total length of the vehicle. In this way, the camera, image processing unit, vehicle type discrimination unit and vehicle type recording unit are housed in a movable housing that is installed on the side of the road, making it easy to carry and install. Is possible.

 According to the method of the present invention, when an axis number counter for counting the number of axes of a passing vehicle is installed on a road, the number of axes counted by the axis number counter is acquired during the passage of the vehicle whose vehicle type is determined. Record according to the discriminating vehicle type. Further, the method of the present invention, when a wheel load meter that measures the wheel load of a passing vehicle is installed on a road, the axle load measured by the wheel load meter during the passage of the vehicle whose vehicle type has been determined. The system is characterized in that the number of axles, axle weight, and gross weight are obtained and the wheel load, the number of axles, axle weight, and the gross weight are recorded according to the discriminated vehicle type. In this case, an alarm is generated when the weight per wheel, per axis, or total weight of the discriminated vehicle exceeds the set value. As described above, the method of the present invention records vehicle type data and other weight data by simultaneously recording the measurement results of the number of axes and the weighing scale installed on the road along with the determined vehicle type. Etc. can be properly combined.

The vehicle monitoring device of the present invention includes a camera that captures a passing vehicle at a predetermined frame period in a movable housing that is installed on the side of a road, and a wheel diameter that is obtained by performing image processing on an image obtained by the camera. An image processing unit that measures the overall length of the vehicle, a vehicle type determining unit that determines the vehicle type based on the wheel diameter measured by the image processing unit and the vehicle overall length, and a vehicle type recording unit that records the vehicle type determined by the vehicle type determining unit. It is characterized by having. The present invention provides a vehicle monitoring method for determining a tire type of a traveling vehicle on a road. This type of tire type discrimination method processes a moving image obtained by capturing the side of a running car with a CCD camera on a frame-by-frame basis, and uses a vector generated by the movement of many feature points in the image to control the vehicle. When a standard template with the least amount of deviation is detected by performing a correlation operation on the tire image of the vehicle and a standard template of various tires corresponding to the shape and size of each vehicle type prepared in advance, And generating a discrimination output of the tire type of the vehicle having the detected standard template. As described above, the present invention determines a tire diameter (type of tire) by preparing a standard template of a tire in advance from a moving image captured by a CCD camera and performing matching, thereby determining the type of vehicle. Can be identified. Also, by identifying the length of the vehicle from the time at which the image of the moving vehicle exists in the moving image, the type of the vehicle can be identified by combining the vehicle length with the tire of the vehicle.

 In this case, the tire type is determined by filling the tire search image and the standard template together, and then comparing the outline of the tire with the change point in the direction opposite to the change point from white pixels to black pixels. By doing so, the tire type is determined. This filter uses a LaBrazian filter. Also, a contour extraction filter such as a Sobel may be used as a filter.

 If the contrast of the image captured at night or the like is low and changes within a certain range, the density distribution is flattened for the frequency distribution of the low-contrast part of the survey image to enhance the contrast of dark parts such as around the tire. And perform a correlation operation. Similarly, when the contrast of an image captured at night or the like is low, the density conversion is performed so as to emphasize only the contrast in the dark part of the search image, and the correlation operation is performed.

In order to accurately determine the tire type, the coordinates of the ground contact point are obtained from the tire image of the vehicle, the lower half image of the tire is cut out, and this is turned upside down to create a template, which is correlated with the upper end of the tire image. To determine the tire type of the search image. The present invention relates to a method for monitoring a vehicle traveling on a road, which process is performed on a frame-by-frame basis for a moving image obtained by imaging a side surface of a traveling automobile with a CCD camera, and is generated by moving a number of feature points in the image. When the vehicle is detected by the vehicle, the tire image of the vehicle is displayed on the tire image of the vehicle, and the tire markers of various types corresponding to the shape and size of each vehicle model prepared in advance. When a standard template with the smallest amount of deviation is detected by performing a correlation operation with the quasi template, a discrimination output of the tire type having the detected standard template is generated for the vehicle, and the vehicle is included in the captured moving image. The speed of the vehicle at the time of detection is calculated based on the distance traveled between frames, and the length of the vehicle is determined using the vehicle transit time from when the vehicle is detected until the image of the vehicle disappears and the speed of the vehicle, The vehicle type is determined using the determined vehicle length and the tire type determination output. Here, the moving amount of each frame of the moving image of the vehicle is integrated, and the vehicle length is calculated in real time in accordance with the speed change of the vehicle based on the integrated value until the image of the vehicle disappears. The vehicle type is determined using the result of the classification.

 The present invention relates to a monitoring device for a vehicle traveling on a road, which stores a CCD camera that captures an image of a side surface of a traveling vehicle, and a standard template of various tires corresponding to the shape and size of each vehicle type. And detecting a vehicle from a motion vector generated by movement of a number of feature points in an image for each frame from a moving image captured by a CCD camera, and detecting an image of a tire of the vehicle and a standard template of the storage unit. A tire discriminator that determines a tire type by performing matching; a speed detector that calculates a speed of the vehicle when the vehicle is detected in the captured moving image based on a moving distance between frames; and a vehicle is detected. A vehicle length determining unit that obtains a vehicle length using the vehicle passing time from when the vehicle image disappears and the vehicle speed, and a vehicle type using the tire type determination output and the determined vehicle length. Characterized in that a vehicle type discriminator for. BRIEF DESCRIPTION OF THE FIGURES

FIG. 1 is an explanatory view of a use state of the present invention;

Figure 2 is an explanatory view of the device configuration of the present invention;

Figure 3 is a block diagram of the image processing pod of Figure 2;

Fig. 4 is an illustration of the processed image when the head of the vehicle enters the center of the screen; Fig. 5 is a characteristic diagram of the gain [m / dot] with respect to the position above and below the screen used for calculating the front wheel diameter;

Figure 6 is an illustration of the processed image when the vehicle rear edge passes through the center of the screen;

FIG. 7 is a flowchart of the vehicle type measurement processing according to the present invention; FIG. 8 is a diagram for explaining another embodiment of the present invention in which detection data of the number of axes are combined;

FIG. 9 is an explanatory view of the discrimination of the vehicle type and the timing of taking in the number of axis count data in the embodiment of FIG.

FIG. 10 is a block diagram of an image processing board used in the embodiment of FIG. 8; FIG. 11 is a flowchart of a vehicle type measurement process according to the embodiment of FIG. 8;

FIG. 12 is an explanatory diagram of another embodiment of the present invention in which detection data of a weighing scale is combined; FIG. 14 is an explanatory diagram of vehicle type discrimination and a loading timing of vehicle weight data in the embodiment of FIG. ;

FIG. 13 is a block diagram of an image processing pod used in the embodiment of FIG. 12; FIG. 15 is a flowchart of a vehicle type measurement process according to the embodiment of FIG. 12; FIG. FIG. 17 is an explanatory diagram showing a screen including a vehicle approach scene;

FIG. 18 is a flowchart of a process for determining a tire in a vehicle image;

Figure 19 is an illustration of tire detection using a standard template;

FIG. 20 is an explanatory diagram of the filtering process;

FIG. 21 is a flowchart of a tire discrimination process when the contrast of an image is low; FIG. 22 is a flowchart of a process for accurately obtaining a tire upper end;

FIG. 23 is an explanatory view showing a specific example of a method of obtaining the upper end of a tire;

FIG. 24 is a block diagram of an image processing unit for determining a vehicle type from a tire type and a vehicle length according to the present invention;

FIG. 25 is a flowchart of a first process for determining a vehicle length;

FIG. 26 is a flowchart of a second process for determining a vehicle length;

FIG. 1 is an explanatory diagram of an installation state of a monitoring device according to the present invention. In FIG. 1, a monitoring device 10 of the present invention is installed on a road shoulder 14 of a roadway 12 and is set so that a field of view 18 of a built-in camera looks down on a road surface. In order to be able to always acquire images under the same conditions with such a monitoring device 10 installed, the roadside 14 must be For example, two white lines are drawn in parallel with the shoulder 14 on the road surface 2 meters and 3 meters away, and the two white lines drawn on the road surface are the marks on the camera monitor provided on the monitoring device 10 Install so that they overlap.

 FIG. 2 is an explanatory diagram of the internal configuration of the monitoring device 10 of FIG. In FIG. 2, a monochromatic CCD camera 20 is provided in the monitoring device 10 of the present invention. The monochromatic CCD camera 20 uses an ultra-wide-angle lens 22 so that the roadway 12 as shown in FIG. The camera has a camera field of view 18 that can sufficiently cover the width of the lane. In addition, a visible light power filter 24 is provided in front of the ultra-wide-angle lens 22, and a near-infrared illumination 26 is further provided. For this reason, the monochrome aperture CCD camera 20 captures an image of a wavelength in the near infrared region in which visible light is reduced by the visible light filter 24. The monitoring device 10 is provided with a power supply unit 28, an image processing pod 30 and an LCD monitor 32. The image processing board 30 has a digital output connector 34 and a serial in connector. 36 are provided.

 FIG. 3 is a block diagram of a circuit configuration of the image processing port 30 provided in the monitoring device 10 of FIG. The image processing port 30 includes an AD converter 38, an image processor 40, an image memory 42, a vehicle type discriminator 44, and a vehicle type recorder 46. The AD converter 38 converts the analog image signal of the near-infrared image captured by the monochrome CCD camera 20 of FIG. 2 into digital gradation data, and stores it in the image memory 42 at predetermined frame periods. To be stored. The image processing unit 40 measures the diameter D of the wheel and the total length L of the vehicle by performing image processing on the image acquired by the monochrome CCD camera. The measurement results of the wheel diameter D and the total length L of the vehicle measured by the image processing unit 40 are given to the vehicle type discriminating unit 44 and compared with predetermined values, for example, for ordinary passenger cars, light trucks, and large vehicles. The vehicle type such as a truck is determined, and the determination result is recorded in the vehicle type recording unit 46.

Next, image processing for measuring the wheel diameter D and the total length L of the vehicle in the image processing unit 40 in FIG. 3 will be described. FIG. 4 is an explanatory diagram of a processing screen for measuring a diameter D of a front wheel of a vehicle. In the processing screen 48, this is an image when the leading F of the vehicle 50 enters the center of the screen, and the leading F of the vehicle enters the center of the screen. At that time, the time T 1 and the processed image 48 are recorded in the image memory 42. In the measurement of the tire diameter D of the front wheel 52 in the processed image 48, the tire diameter D is measured based on the number of dots in the height direction of the front wheel 52 on the screen, for example, Ν dot. Here, in the monitoring apparatus 10 of the present invention, the camera view 18 is set so that the roadway 12 looks down on the road surface from the shoulder 14 as shown in FIG. The lower the screen above, the closer the distance to the monitoring device 10, and the higher the position above the screen, the farther the distance from the monitoring device 10. For this reason, the number of dots in the height direction corresponding to the tire diameter D of the front wheels 52 increases even if the tire diameter D does not change. However, the closer to the centerline, the farther the tires appear, the smaller the number of dots. Therefore, it is necessary to change the sensitivity for determining the actual dimensions around one dot according to the vertical position of the front wheel 52 on this screen.

 FIG. 5 is a characteristic diagram showing a relationship between the number of dots i from the upper side of the image to the lower side on the processing screen 48 of FIG. 4 and a gain G [m / dot] indicating an actual length per dot with respect to the number of dots i. . As is clear from this characteristic diagram, the smaller the number of dots from the upper side, that is, the greater the number of dots from the upper side, that is, the greater the number of dots from the upper side, that is, the lower the number of dots from the upper side of the screen. In other words, the gain G becomes smaller. Therefore, in the processed image 48 of FIG. 4, for example, if the number of dots from the upper part of the screen to the lower end of the front wheel 52 is i, the gain G i is plotted by plotting the number i of dots as shown in FIG. You can ask. Once the gain G i is obtained in this way, the tire diameter D can be calculated by multiplying the gain by the N dot indicating the height of the tire in the theoretical image 48 in FIG.

 D = G i X N

Can be calculated as The gain G i in this case is obtained from i dots, which is the number of dots from the upper end of the image to the lower end of the front wheel 52, but the tire is in the region from i dot to (i—N) dots. In this case, for example, By converting the N dots to the tire diameter D using the average value of the gains of the i dot and (i-N) dots, the tire diameter can be obtained with higher accuracy. Like this If the tire diameter D of the front wheels 52 can be calculated from the image when the head F of the vehicle comes to the center of the screen, then the vehicle passes and the rear end R of the vehicle is processed as shown in Fig. 6. The time T2 and the image when passing through the center of the vehicle are stored, and the vehicle is traveling from the image of the front part F of the vehicle at the time T1 in FIG. 4 to the image of the vehicle rear end R in the processed image 48 in FIG. The vehicle length L is obtained by integrating the movement of the vehicle for each frame. To detect the movement of the vehicle in each frame, a difference image is obtained by subtracting the image of the current frame from the image of the previous frame, and the number of dots indicating the width of the movement in the traveling direction of the vehicle appearing in the difference image is determined for each frame. In this case, the number of pixels indicating the movement of the vehicle in each frame can be calculated by calculating the gain G in Fig. 5 corresponding to the number of dots i from the upper side of the image. By multiplying the direction, that is, the number of pixels in the horizontal direction, the amount of movement of the vehicle for each frame can be calculated. FIG. 7 is a flow chart of the vehicle type measurement processing of the present invention using the image processing port 30 of FIG. First, in step S1, it is checked whether or not the head F of the vehicle enters the center of the screen as shown in the processing image 48 of FIG. When the head F of the vehicle enters the center of the screen, the process proceeds to step S2, and the time T1 and the image are recorded. Subsequently, in step S3, the diameter D of the front wheel 52 is calculated from the recorded image. Subsequently, in step S4, it is checked whether or not the rear end R of the vehicle passes through the center of the screen as shown in a processed image 48 of FIG. Until the rear end R of the vehicle passes through the center of the screen, a frame image is recorded in step S5. When the rear end R of the vehicle passes through the center of the screen, the process proceeds to step S6, and the time T2 and the image are stored. Subsequently, in step S7, the vehicle length L is calculated by integrating the motion amount of each frame during the passage of the vehicle. After the front wheel diameter D and vehicle length L of the passing vehicle are calculated in this way, the type of vehicle is determined from the front wheel diameter D and vehicle length L in step S8. In this embodiment, a normal passenger car in step S9, a normal truck in step S10, or a large truck in step S11 is determined by using a predetermined determination value. If the determination result in step S9 or step S11 is obtained, the vehicle type determination result is recorded in step S12, and the vehicle type measurement for the passing vehicle is completed. Subsequently, in step S13, the presence or absence of a stop instruction is checked. If there is no stop instruction, the flow returns to step S1 again, and the same processing is performed for the next passing vehicle. Repeat the process.

 FIG. 8 shows another embodiment of the monitoring device according to the present invention. In this embodiment, the measurement results of the axis number counter 54 separately installed in the monitoring device 10 of the present invention are combined. It is characterized by the following. In FIG. 8, a monitoring device 10 of the present invention is installed on the shoulder of the road in the same manner as shown in FIG. In this embodiment, in addition to the monitoring device 10 according to the present invention, an axis number counter 54 is provided. The number-of-axes counter 54 is provided with, for example, a pressure-sensitive mat 56 on the roadway 12, and is installed so that the left wheel of the vehicle 17 passes over the pressure-sensitive mat 56. The axis number counter 54 receives an axis number count-up input each time a wheel passes through the pressure-sensitive mat 56 and counts up the counter. The monitoring device 10 of the present invention, when determining the vehicle type of the passing vehicle, captures the axis count-up input measured by the axis counter 54 during the passage of the vehicle, so that the passing vehicle that has performed the vehicle type determination is obtained. Determine the number of axis data and record it along with the vehicle type data.

 FIG. 9 is a timing chart of data acquisition from the axis number counter 54 in the monitoring apparatus 10 of FIG. For the number of axes counter 54, the pressure-sensitive mat

Each time the wheel of the vehicle passes on 56, the axis count-up input 58a, 58b, 58c is obtained. Specifically, the axis number count-up inputs 58a to 58c and the input times are stored. The monitoring device 10 records the vehicle entry time T1 and the vehicle rear end passage time T2 for the image captured by the camera when the vehicle 17 passes. Therefore, of the axis count-up input 58-58c taken from the axis count counter 54, the axis count-up input existing between the vehicle entry time T1 and the vehicle rear end time T2 Based on 58b and 58c, it detects that the number of passing vehicle axes is two, and combines and records the data for the discriminating vehicle type.

 FIG. 10 is a block diagram of the image processing board 30 provided in the monitoring device 10 of the embodiment in FIG. In this image processing port 30, the AZD conversion section 38, image processing section 40, image memory 42, vehicle type discriminating section 44, and vehicle type recording section provided on the image processing board 30 of FIG. 4 In addition to 6, new axis number data processing section

60 is provided. The number-of-axes data processing unit 60 determines the type of passing vehicle When the vehicle type is determined, the count information from the number-of-axes counter 54 at that time is taken in, and as shown in FIG. 9, the axis input between the vehicle entry time T1 and the vehicle rear end passage time T2 is obtained. The number of axes of passing vehicles is obtained from the number of the number count-up inputs, and the number of axes data is recorded in the vehicle type recording section 46 together with the data of the discriminated vehicle type by the vehicle type discriminating section 44.

 FIG. 11 is a flowchart showing the processing of the monitoring apparatus 10 that combines the data of the axis number counter 54 of FIG. The vehicle type discriminating process in step S1 has the processing contents of steps S1 to S12 in FIG. 7, and the vehicle type is discriminated by image processing and recorded. Subsequently, in step S2, the number of axes counted while passing through the vehicle is fetched, and in step S3, the result of the vehicle type determination and the measured number of axes are recorded. Then, the processing from step S1 is repeated until a stop instruction is issued in step S4. FIG. 12 shows another embodiment of the monitoring device according to the present invention. In this embodiment, the vehicle weight data of the wheel load meter 62 separately installed in the monitoring device 10 of the present invention is taken. It is characterized by combining with. In the embodiment of FIG. 13, a weighing machine 62 is separately installed in addition to the monitoring device 10 according to the present invention. The weighing machine 6 2 is provided with a sheet-shaped load sensor 6 4 at a position where the left wheel of the vehicle 17 passing through the roadway 12 passes, and the load sensor 6 4 when the wheel of the passing vehicle passes. The wheel weight added to the is measured. Each time the monitoring device 10 determines the type of passing vehicle, the monitoring device 10 captures the data of the wheel load meter 62, obtains the number of axles, wheel loads, axle loads, and gross weight, and records the data together with the vehicle type data.

FIG. 13 is a time chart of the timing of capturing the vehicle weight data of the wheel weighing machine 62 by the monitoring device 10 of FIG. In the wheel load meter 62, each time a vehicle wheel passes through the weight sensor 64 installed on the roadway 12, the vehicle weight data 6 The wheel weight, which is the weight per wheel, is measured. When the monitoring device 10 determines the type of the passing vehicle, the monitoring device 10 takes in the measurement data of the weighing machine 62 and obtains and records the data concerning the weight of the vehicle. Specifically, the monitoring device 10 measures the vehicle entry time T1 and the vehicle rear end time T2 as shown in the timing chart of Fig. 13 by image processing of passing vehicles. , The vehicle measured between the vehicle entry time T 1 and the vehicle rear end passage time T 2 Weight data input 6 8b and 6 8c are taken in, and the number of axes, axle weight, and total weight are calculated. That is, the number of axes is the number of vehicle weight data inputs 68b and 68c between T1 and T2, that is, the number of axes is 2. The axle weight was doubled for each wheel weight based on vehicle data input of 68b and 68c.

 Axle load = wheel load X 2

It is. In addition, the total weight of the vehicle is the sum of the front wheel axle weight and the rear wheel axle weight

 Gross vehicle weight = front wheel axle weight + rear wheel axle weight

Is obtained as Further, in the monitoring device 10 of the embodiment of FIG. 12, the number of axles, wheel loads, axle loads, and gross weights detected from the data taken by the weighing machine 62 in accordance with the vehicle type discrimination of the passing vehicle are described. A comparison is made with a predetermined value, and if any of the values exceeds a predetermined value, an alarm signal is output to an alarm device 66 installed separately to generate an alarm. Thus, for example, an alarm can be issued to a passing vehicle or the like that is overloaded to notify the driver.

 FIG. 14 is a block diagram of the image processing port 30 provided in the monitoring device 10 in the embodiment of FIG. In this image processing port 30, an A / D conversion section 38, an image processing section 40, an image memory 42, and a vehicle type discriminating section 4 provided in the image processing port 30 of FIG. 4 and a vehicle type recording unit 46, and a vehicle weight data processing unit 70 is provided. The vehicle weight data processing unit 70 captures the vehicle weight data input from the wheel load meter 62 in Fig. 12 as shown in the timing chart of Fig. 3 when the vehicle type discriminating unit 44 determines the vehicle type of the passing vehicle. The number of axles, wheel weights, axle weights, and gross weights are determined based on the vehicle weight data input between the time of entry T1 at the head of the vehicle and the time of passage T2 at the rear end of the vehicle. The data is combined with the vehicle type data recorded. In the vehicle data processing unit 70, the number of axles, wheel loads, axle loads, and gross vehicle weight obtained from the measurement results of the weighing scale are compared with predetermined values, and an alarm is issued if the values exceed the predetermined values. It will output a signal to trigger an alarm.

FIG. 15 is a flowchart of the processing by the monitoring apparatus 10 of the present invention in the embodiment of FIG. The vehicle type determination processing in step S1 is the same as the vehicle type determination by the image processing in steps S1 to S12 shown in FIG. After the vehicle type determination is completed, the data measured during the passage of the vehicle from the weighing machine 62 in step S2 is obtained. And calculate the number of axles, wheel loads, axle loads, and total weight. Subsequently, in step S3, it is checked whether any of the number of axles, wheel load, axle load, and gross weight exceeds a predetermined value, and if so, an alarm signal is output in step S4, An alarm is issued by an alarm. Subsequently, if there is no stop instruction in step S5, the process returns to step S1 again, and the same processing is repeated.

 FIG. 16 is a block diagram of a processing function for the tire type determination of the present invention by the image processing unit in FIG. In FIG. 16, the image processing unit 40 includes a vehicle image detection unit 72, a tire determination unit 74, and a standard template storage unit 76. The standard template storage section 76 stores standard templates for tires having diameters corresponding to the types of vehicles such as large vehicles, medium vehicles, small vehicles and ordinary vehicles. The diameter of the tire is about 100 Omm for the large model, about 900 mm for the medium model, and about 600 mm to 700 mm for the small model. It stores a standard tire template using the captured image.

 As shown in Figs. 1 and 2, a vehicle passing through the roadway is photographed by the CCD camera 20 which is also received by the monitoring device 10 installed on the shoulder 14 near the roadway 12 and a moving image is shown. Input to the image processing section 40 of 16. Here, when a moving vehicle is included in the image, the vehicle is detected as a vehicle by detecting a vector based on the movement of a feature point in each frame period captured by the vehicle image detecting unit 72. As a result, the tire discriminating unit 74 is driven, the standard templates are taken out one by one from the tire standard template storage unit 76, matched with the tire image in the vehicle image, and the degree of mismatch is detected. As a result of matching with the standard template, if the standard template with the least degree of mismatch is identified, the tire type (evening diameter) of the standard template is extracted and output.

FIG. 17 shows a frame image including a vehicle image of a vehicle approach scene. The frame image 78 taken by the CCD camera 20 is subjected to processing such as correlation calculation in the image processing unit 40. For the frame image 78, for example, a correlation operation area 80 0-11 to 80-56 of 5 (horizontal) x 6 (vertical) = 30 small areas is formed. . The frame period of the frame image 78 captured by the CCD camera 20 is 1Z30 seconds, and the current frame and the image of the previous frame are compared. Find the momentum. When the head of the vehicle enters the photographing range of the CCD camera 20, the movement amount, that is, the movement is detected from each correlation calculation result, and the vehicle entry can be detected. Correlation calculation area 80 0-11 to 8 0-56 The contents of the correlation calculation are performed for each frame, and the characteristic points of the vehicle, such as car window frames, doors, mirrors, and tires, are continuously parallel for each frame. It is detected as a vector that moves to the vehicle, and it is possible to identify the approach of the vehicle. Garbage other than cars is excluded because the vector cannot be detected continuously over a wide range by the correlation operation.

 FIG. 18 is a flowchart of a process for determining a tire in a vehicle image. First, in step S1, a vehicle image is obtained by performing a correlation operation on the image of the correlation calculation area 80-11 to 80-56 set in the frame image 78 input as shown in Fig. 17 Is done. Next, the process proceeds to step S2, where it is determined whether or not the matching ratio of the tire image of the vehicle image with the previously prepared standard template is equal to or greater than a predetermined value. In this case, the matching rate is high when the degree of mismatch (area) between the tire image (search image) of the vehicle image and the standard template is small. Therefore, if the matching ratio is equal to or greater than a predetermined value, the size (type) of the tire is specified in step S3. If the matching ratio is less than the predetermined value, it is determined in step S4 that no evening is detected.

FIG. 19 is an explanatory diagram of a tire detection method. In Fig. 19, the frame images are changed in the order of 78-1, 78-2, and 78-3, showing that the vehicle is moving from right to left on the screen, several frames from the time of entering the vehicle. Save the frame image of. For the saved frame images, the tires are searched for using the standard templates corresponding to a plurality of types classified according to the shapes and sizes prepared in advance. For the tires in this vehicle, as a standard template, a circular pattern 841-1 to 84-3, in which all tires and wheels are composed of black pixels, as shown in pattern group 84, and a pattern group A pattern with a white pixel on the wheel at the center of the circle, which is a tire taken with a standard lens as shown in Figure 86, and a wide-angle lens as shown in Pattern Group 88 with a white pixel on the wheel. The patterns 88-1 to 88-3, in which the white pixel of the wheel is provided at the center of the oval shape of the tire, are prepared. This example shows only three sizes for three patterns, but in practice more Many patterns and many sizes are available.

 In the example of FIG. 19, it is shown that the tire image 75 of the frame image 78-3 is matched with the pattern 84-1 of the standard template. Since this standard template corresponds to the actual tire, the diameter of the tire in the image can be known from the matched standard template. Here, the conversion from the image coordinates to the actual dimensions uses a conversion formula determined according to the coordinates of the ground contact point of the tire. Further, the standard template may be an actual tire image or an image after filtering such as Laplacian. In other words, by applying a filter to both the tire search image on the screen and the standard template tires, the contours that represent the characteristics of the car, that is, changes from white pixels to black pixels and vice versa, are noticed. Processing, so it is not easily affected by the brightness of the image. In addition, the possibility of matching can be increased by blurring the contour with a filter.

 FIG. 20 is an explanatory diagram of the filtering process. Figure 20 (A) shows the tire image, Figure 20 (B) shows the pixel distribution when the center point of the tire image is scanned horizontally, and Figure 20 (C) shows an example of the Laplacian transform of the tire image. It has peaks of positive and negative polarities indicating the boundary of the contour line, and can identify a change from a white pixel to a black pixel and a change from a black pixel to a white pixel. FIG. 20 (D) shows an example of the Sobel transform of a tire image, in which only peaks representing contours are shown. When discriminating vehicles at night or the like, when the image of the vehicle captured by the CCD camera has low contrast and the occurrence distribution of dark areas such as around the tires is large, the image density distribution is flattened and darkened. Performs image conversion processing that emphasizes the contrast of the part, and performs correlation calculation with the standard template. Further, when the contrast is low at night or the like, the dark area data may be expanded so as to emphasize only the contrast of the dark area, and the tire in the image may be extracted by performing the correlation operation.

FIG. 21 is a flowchart of the tire discriminating process for the case where the contrast of the image is low. When a vehicle image is acquired in step S1, it is determined in step S2 whether the contrast of the image is equal to or less than a threshold. If the contrast is equal to or less than the threshold, flattening processing of image density or extension of dark area data is performed in step S3. Subsequently, in step S4, it is determined whether the matching rate with the standard template is equal to or greater than a predetermined value. Judge that there is no tire and return to step SI. If the matching ratio is equal to or larger than the predetermined value in step S4, the size of the tire is specified in step S6 from the standard template at that time.

 When performing such tire discrimination processing, the boundary between the fender of the car and the upper part of the tire may not be clear depending on the type of the car or the observation time in the tire image. In order to cope with such a case, the present invention uses a method of creating an upper half template using an image of a lower half of a tire whose contour can be relatively clearly identified.

 FIG. 22 is a flowchart of a process for accurately obtaining the upper end of the tire, and FIG. 23 is a specific example of a method for obtaining the upper end of the tire. Note that the image 90 in FIG. 23 represents the whole of the taken tire, in which the upper half is clearly shown, and the outline is actually not clear due to the fender.

 First, a vehicle image is acquired in step S1, and it is determined in step S2 whether the matching ratio with the standard template is equal to or greater than a predetermined value. If the matching ratio is not equal to or more than the predetermined value, the process proceeds to step S3, where it is determined that there is no tire, and the process returns to step S1. If the matching ratio is equal to or more than the predetermined value, the lower end of the tire is found in step S4. In this case, the lower end of the evening is detected based on the difference in color between the pixels on the road and the pixels on the tire contact portion. In the example of Fig. 23, the lower half of the lower end side of the tile image 90 composed of the upper half image 90-1 and the lower half image 90-2 is detected and cut out as the image 90-3. You. Next, in step S5, the image of the lower half of the tire is turned upside down to create a template. In the example of FIG. 23, an upside down template 90-4 is obtained by inverting the lower half image 90-3. Subsequently, in step S6, the upper end of the tire is found using the upside down template. At this time, in the example of FIG. 23, matching is performed between the upside down template 90-4 and the upper half image 90-1 of the original tire image 90, and the matching ratio is equal to or higher than a predetermined value. Is obtained, it means that the upper end of the tire has been found. Thereafter, a process of specifying the size of the tire by matching the tire image having the upper end found in step S7 with the standard template of each tire is performed.

Figure 24 shows the vehicle length detected based on the tire discrimination. FIG. 4 is a block diagram of a functional configuration of an image processing unit according to the present invention for determining a type. The image processing unit 40 includes a vehicle length detection unit 92 and a vehicle type determination unit 94 in addition to the vehicle image detection unit 72, the tire determination unit 74, and the standard template storage unit 76 shown in FIG. Is provided.

 In this embodiment, the tire discriminating unit 74 is driven when the vehicle image is detected by the vehicle image detecting unit 72 for a moving image captured by the CCD camera 20 at each frame period by detecting a vector by moving a feature point. Then, the standard templates are taken out one by one from the tire standard template storage unit 76 and matched with the tire image in the vehicle image, the degree of mismatch is detected, and as a result of matching with all standard templates, By identifying the standard template with the least degree of mismatch, the tire diameter (type) corresponding to the standard template is output as the identification result. In some cases, the vehicle type cannot be accurately determined only by the discrimination result of the evening diameter, and the vehicle length determining unit 92 is driven to accurately determine the vehicle type. There are several principles of vehicle length discrimination, one of which can be determined by the product of the speed at which a video of a vehicle is detected and moving on the screen and the time from when the vehicle enters to when it disappears. Thus, the vehicle type can be determined by the vehicle type determining section 94 based on the identification result of the tire diameter (type of tire) and the vehicle length. Next, there are two methods for determining the length of the vehicle, some of which differ in principle. The processing flow for implementing each method will be described with reference to FIGS. 25 and 26.

FIG. 25 is a flowchart of a first process for detecting a vehicle length. This process is started for each frame of the image. In step S1, it is first determined whether or not there is a motion vector equal to or larger than the threshold in the screen. If there is motion, the process proceeds to step S2, and the car is moved one frame earlier. Determine if there is nothing. Here, if it is found that there was no vehicle one frame before, the process returns to step S1, but if there is a vehicle, it is determined that the vehicle has entered and the process proceeds to step S3, where the speed at the time of entry is reduced by the movement per frame. Calculate and acquire based on the number of pixels, and return to step S1. After the vehicle has been detected and its speed has been detected in this way, when the vehicle passes in front of the CCD camera, it is determined in step S1 that there is no motion vector above the threshold in the screen, and step S4 Move to Here, it is determined whether or not there is a vehicle one frame before.If there is no vehicle, the process returns to step S1.If there is, the process goes to step S6 after recognizing completion of the vehicle in step S5, and proceeds to step S6. The time from when the vehicle enters the vehicle image until the vehicle disappears The length of the vehicle is determined from the product of the speed at the time of approach.

 FIG. 26 is a flowchart of the second process for determining the vehicle length. This process is also started for each frame as in FIG. 25, and steps S1 and S2 are executed similarly. In step S2, when it is determined that the vehicle is present one frame before, the speed at that time, that is, the movement amount for each frame is integrated in step S4. By this integration, the movement amount up to that point, that is, the length, is obtained for each frame. If it is determined in step S2 that there is no vehicle one frame before, it is determined that the vehicle is approaching as in step S3 in FIG. 25, the speed at that time is obtained, and the process returns to step S1. If it is determined in step S1 that there is no motion vector equal to or larger than the threshold value on the screen, steps S5 and S6, which are the same processes as steps S4 and S5 in FIG. 25, are executed. That is, when it is determined in step S1 that there is no motion vector, the process proceeds to step S5, and if there is a vehicle one frame before, it is recognized that the vehicle has passed in step S6, and in step S7 The accumulated amount up to now is defined as the vehicle length. In the second process of FIG. 26, a relatively accurate vehicle length can be obtained, for example, even when the driver of the car applies a brake.

 In step S1 in Figs. 25 and 26, when detecting the motion vector of the vehicle, the result of detecting the vector of a featureless place such as a door has low reliability. It is desirable to use only the peak of the contour whose difference from the peak of the large contour is equal to or larger than a predetermined value. If there is no feature point during the passage of the vehicle and a valid vector cannot be obtained, interpolation is performed using the values of the preceding and following frames or the average value. In addition, the length of the vehicle thus obtained can be actually obtained by using a conversion formula corresponding to the coordinates of the ground contact point.

In the above embodiment, as shown in FIG. 2, the monitoring device 10 of the present invention alone measures and records the type of passing vehicle by image processing, as shown in FIG. In addition, since the monitoring device 10 is provided with a digital output connector 34 and a serial interface connector 36, the measurement results can be recorded and monitored remotely using these output units and the interface. You can also. In addition, the present invention is not limited to the above embodiment, and includes appropriate modifications that do not impair the objects and advantages thereof. Industrial applicability

 As described above, according to the present invention, it is possible to continuously and accurately determine the type of a vehicle traveling on a road surface at an arbitrary road shoulder, and it is possible to collect vehicle type data without any person. Further, a vehicle type discrimination system using image processing can be realized. The camera image processing unit, vehicle type discrimination unit, and vehicle type recording unit are housed in a movable housing installed on the side of the road, making it easy to carry and install, and classifying and counting vehicle types at any location. Can be done unattended.

 Also, the measurement data of the axis counter and weight meter installed at the same location can be taken and stored together with the discriminated vehicle type, and if necessary, an alarm can be issued when the weight data exceeds the specified value. In addition, it is possible to collect and record more detailed data of passing vehicles required for road management and the like unmannedly and accurately.

Claims

The scope of the claims 1. In the method of monitoring vehicles traveling on the road,  An image of a passing vehicle is taken at a predetermined frame cycle by a camera installed in a movable housing that is installed on the side of the road.  Image processing of the image acquired by the force mera measures the diameter of the wheels and the total length of the vehicle,  It is characterized in that the vehicle type is determined and recorded from the measured wheel diameter and the total length of the vehicle. 2. In the method according to claim 1, when an axis number counter for counting the number of axes of vehicles passing therethrough is installed on the road, the axis power input is performed during the passage of the vehicle whose vehicle type is determined. The feature is that the number of axes counted in the evening is taken in and recorded according to the discriminated vehicle type. 3. In the method according to claim 1, when a wheel load meter for measuring the wheel weight of the passing vehicle is installed on the road, the wheel load meter measures the vehicle during the passage of the vehicle whose type is determined. The obtained axle load is taken in, the number of axles, the axle weight and the gross weight are obtained, and the wheel load, the number of axles, the axle weight and the gross weight are recorded according to the discriminated vehicle type. 4. The method according to claim 3, wherein an alarm is generated when the weight per wheel, the weight per axle, or the total weight exceeds the set value. 5. A mobile monitoring device installed on the side of the road to monitor vehicles running on the road,  A camera that images a passing vehicle at a predetermined frame cycle; An image processing unit that measures the diameter of the wheel and the total length of the vehicle by performing image processing on the image acquired by the camera; A vehicle type determining unit that determines a vehicle type from the wheel diameter and the total vehicle length measured by the image processing unit;  A vehicle type recording unit that records the vehicle type determined by the vehicle type determination unit, Are provided integrally. 6. In the device according to claim 5, when the number of axes counting the number of axes of the passing vehicle is set on the road, the number of axes counting is set during the passage of the vehicle whose vehicle type is determined. An axis number data processing unit is provided which takes in the number of axes counted in step (1) and records it in accordance with the discriminated vehicle type. 7. In the device according to claim 5, when a wheel load meter for measuring the wheel weight of the passing vehicle is installed on the road, the wheel load meter measures the vehicle during the passage of the vehicle whose type is determined. A vehicle weight data processing unit for obtaining the number of axles, axle weight and gross weight by taking in the obtained axle weight, and recording the wheel load, the number of axles, the axle weight and the gross weight according to the discriminated vehicle type. And 8. The apparatus according to claim 7, wherein the vehicle weight data processing unit generates an alarm when the weight per wheel, the weight per axle, or the total weight of the discriminated vehicle exceeds a set value. It is characterized by the following. 9. In the method of monitoring vehicles traveling on the road,  A moving image captured by a CCD camera of the side of a running car is processed for each frame, and when a vehicle is detected by a vector generated by the movement of many feature points in the image, a tire image of the vehicle is prepared in advance. When the standard template with the smallest amount of deviation is detected by performing correlation calculation with the standard template of each type of tire corresponding to the shape and size of each vehicle type, the vehicle is held with the detected standard template. It is characterized in that a tire type discrimination output is generated. 0. The method of claim 1 wherein the tire search image and the standard After applying the fill pattern to the tires, the tire type is determined by comparing the outline of the tire with the change point in the direction opposite to the change point from the white pixel to the black pixel. 11. The method according to claim 10, wherein a Laplacian filter is used as the filter. 12. The method according to claim 10, wherein a contour extraction filter such as a Sobel is used as the filter. 13 3. In the method of claim 9, when the contrast of the image taken at night or the like is low and changes within a certain range, the density distribution is flattened in the frequency distribution of the portion where the contrast of the exploration image is low. The correlation calculation is performed by emphasizing the contrast in dark areas such as around the tires. 14. According to the method of claim 9, when the contrast of an image captured at night or the like is low, the density conversion is performed so as to emphasize only the contrast of the dark part of the search image, and the correlation calculation is performed. Features. 15. In the method according to claim 9, the coordinates of the ground contact point are obtained from the tire image of the vehicle, an image of the lower half of the tire is cut out, and the image is inverted upside down to create a template. It is characterized by accurately determining the tire type of the search image by correlating with the upper end of the search image. 1 6. In the method of monitoring vehicles traveling on the road, A moving image captured by a CCD camera of the side of a running car is processed for each frame, and when a vehicle is detected by a vector generated by the movement of many feature points in the image, a tire image of the vehicle is prepared in advance. By calculating the correlation with the standard template of each type of tire corresponding to the shape and size of each model, the amount of deviation can be reduced. When the least standard template is detected, a discrimination output of the type of the tire having the detected standard template is generated for the vehicle,  Calculating a speed of the vehicle when the vehicle is detected in the captured moving image based on a moving distance between frames;  A vehicle length is obtained using a vehicle transit time from the detection of the vehicle to the disappearance of the image of the vehicle and the speed of the vehicle,  It is characterized in that the vehicle type is determined using the determined vehicle length and the determination output of the tire type. 17. The method according to claim 16, wherein the moving amount of each frame of the moving image of the vehicle is integrated, and a real-time corresponding to a change in the speed of the vehicle based on the integrated value until the image of the vehicle disappears. Means for determining the vehicle length, and the vehicle type is determined using the determined vehicle length and the tire type result.  1 8. In the monitoring device of vehicles running on the road,  A C C D camera that images the side of the running car,  A storage unit for a standard template for each type of tire corresponding to the shape and size of each vehicle type,  A vehicle is detected from a motion vector generated by movement of a number of feature points in an image for each frame from a moving image captured by the CCD camera, and matching of an image of the tire of the vehicle with a standard template in the storage unit is performed. And an evening discriminator for discriminating the tire type by taking  A speed detecting unit that calculates a speed of the vehicle when the vehicle is detected in the captured moving image based on a moving distance between frames;  A vehicle length determining unit that determines a vehicle length using a vehicle transit time from when the vehicle is detected to when the image of the vehicle disappears and a speed of the vehicle;  A vehicle type determination unit that determines a vehicle type using the tire type determination output and the determined vehicle length, It is characterized by having.