JPH11259792A - Method and device for recognizing vehicle - Google Patents

Abstract

PROBLEM TO BE SOLVED: To detect a specific type of a vehicle easily and also with high accuracy by executing detection processing with a characteristic representing the front of this type of a vehicle at the time of detecting a specific type of a vehicle by two-dimensional image processing and three-dimensional measurement processing. SOLUTION: A large vehicle detecting part calculates the distances to respective models MF about a projection point that is located in the neighborhood of the projection models MF at that point of time for every scan position, further calculates the mean value of each distance and makes it a difference degree between a projection image represented by each projection point and the model MF. The difference degree of each scan position is compared with a prescribed threshold and when the difference degree is below the threshold and also a scan position ZA that becomes the minimum value is detected, a CPU recognizes the position ZA as the front position of a large vehicle. Also, the CPU recognizes each projection point included in an area from the position ZA to a place ZB that is behind only by a prescribed distance d as a projection point of a configuration point of the large vehicle and excludes it from an object of matching processing.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a technique for recognizing a vehicle located on a road, a parking lot, or the like. A method and apparatus for recognizing a vehicle located in an area.

[0002]

2. Description of the Related Art Heretofore, there has been an apparatus which is provided with a television camera at a position near a road, processes images from the camera, recognizes vehicles on the road, and measures traffic flow such as the number of vehicles passing and the speed of passage. Exists. In this device, first, after a background image or an image obtained a predetermined time ago is subtracted from the image obtained by the camera, a binarization process or an edge extraction process is performed on the image after the difference processing, A change area of the image data is extracted. Each vehicle is separated and detected by scanning a two-dimensional model of the vehicle on the extraction result. Furthermore, by associating the detection results of vehicles along the time axis and tracking changes in the detection position of each vehicle,
The number and speed of vehicles passing can be calculated.

Recently, the applicant has proposed a vehicle detection method using a three-dimensional recognition and pattern matching technique in order to detect a vehicle on a road with high accuracy (Japanese Patent Application Laid-Open No. Hei 9 (1994)).
-33232). In this method, first, after an observation region is imaged by a plurality of cameras, feature points of the vehicle extracted on each image are associated with each other between the images, and three-dimensional coordinates of an object point indicating the vehicle are measured. You. Thereafter, each three-dimensional coordinate is projected on a virtual vertical plane along the road, and the side shape of each vehicle is extracted. Then, a plurality of types of vehicle projection models are sequentially scanned on the projection image, and each vehicle at the observation position is separated and recognized.

[0004]

In the case of performing vehicle recognition by extracting a change area of image data on a two-dimensional image, if the change area of the image data of each vehicle is continuous due to traffic congestion or the like, a plurality of changes occur. There is a risk that one general vehicle will be erroneously recognized as a large vehicle.

[0005] In the case of a large vehicle, the shape and length of the rear part of the vehicle body differ depending on the vehicle, so that the shape of the image area corresponding to these vehicles also varies. As a result, it is very difficult to accurately recognize a large vehicle, for example, it is recognized that the image regions of a plurality of general vehicles are continuous even though the image region is a large vehicle.

In the method of detecting a vehicle using the result of projecting a three-dimensional measurement result on a virtual plane, a large vehicle may be erroneously detected due to the difference in the shape and length of the rear portion of the vehicle body. high. In order to prevent this erroneous detection, it is necessary to set a plurality of types of projection models for a large vehicle, and sequentially scan each projection model on a virtual plane to perform a matching process. However, when the number of projection models is increased, the capacity of the memory for storing the models is increased, and inconveniences such as a long time required for vehicle detection and an increase in the cost of the apparatus are caused.

SUMMARY OF THE INVENTION The present invention has been made in view of the above-mentioned problems. When a vehicle of a specific vehicle type is detected by using an image obtained by an image pickup means or a three-dimensional measurement result in an observation area, the vehicle is used. It is an object of the present invention to efficiently and highly accurately recognize a vehicle by searching for a feature corresponding to the front of a vehicle.

[0008]

For example, the size and shape of the front of a large vehicle such as a truck or a bus differ greatly from those of a general vehicle, so that an image area corresponding to the front of the large vehicle or a three-dimensional image is obtained. If the measurement data is searched, a large vehicle can be accurately detected. Also, since the front part such as the windshield takes a similar shape and size if the vehicle type is the same, by detecting the image area and the three-dimensional measurement data corresponding to the front part of the vehicle of the specific model, the rear part is obtained. Regardless of the shape or length, a vehicle of a specific vehicle type can be easily detected.

The present invention has been made on the basis of the above principle. According to the first aspect of the present invention, a vehicle located in an observation area is obtained by using an image obtained by imaging a predetermined observation area by an imaging means. In the method of recognizing, the image area corresponding to the front of the vehicle of the specific vehicle type is searched on the image, and when the corresponding image area is detected, the vehicle of the specific vehicle type is located at a spatial position corresponding to the detected position. I try to recognize that it exists.

In the vehicle recognition method according to the second aspect of the invention, a matching process is performed by scanning a model image representing the front of a specific vehicle type on the image, and when an image area corresponding to the model image is extracted, The system recognizes that the vehicle of the specific vehicle type exists at a spatial position corresponding to the extraction position.

According to the third and fourth aspects of the present invention, a three-dimensional measurement process is performed in the observation area using an image obtained by imaging a predetermined observation area by at least one image pickup means. The present invention relates to a method for recognizing a vehicle located in an observation area by using a result. The invention according to claim 3 searches for a measurement result corresponding to the front of a vehicle of a specific vehicle type from the three-dimensional measurement results, and When the measurement result is detected, it is recognized that the vehicle of the specific vehicle type exists at the detection position.

According to a fourth aspect of the present invention, after projecting each three-dimensional coordinate measured by the three-dimensional measurement processing onto a virtual plane positioned perpendicular to the bottom surface of the observation area, the specific vehicle type of the specific vehicle type is projected on this virtual plane. The front projection model is scanned, and when a projection image corresponding to the model is extracted, it is recognized that a vehicle of the specific vehicle type exists at a spatial position corresponding to the extraction position.

According to a fifth aspect of the present invention, there is provided a vehicle recognition apparatus comprising:
Image input means for inputting an image obtained by imaging a predetermined observation area by an image pickup means, and search means for searching an image input by the image input means for an image area corresponding to the front of a vehicle of a specific vehicle type And a recognizing means for recognizing that when the image area is detected by the searching means, an image of the vehicle of the specific vehicle type is present at the detected position.

According to a sixth aspect of the present invention, there is provided a vehicle recognition device comprising:
Image input means similar to the above, matching processing means for scanning a model image representing the front of a specific vehicle type on an image input by the image input means, and performing matching processing, and the model image by the matching processing And a recognizing means for recognizing that an image of the vehicle of the specific vehicle type exists at an extraction position of the image area when the image area corresponding to the image area is extracted.

According to a seventh aspect of the present invention, there is provided a vehicle recognition apparatus comprising:
Image input means for inputting an image obtained by imaging a predetermined observation area by at least one image pickup means, and three-dimensional measurement processing in the observation area is performed using the image input by the image input means. Three-dimensional measuring means
Searching means for searching the three-dimensional measurement result for measurement data corresponding to the front of a vehicle of a specific vehicle type, and when the measurement data is detected by the searching means, the detected position is set to the top of the vehicle of the specific vehicle type And recognition means for recognizing the position.

The vehicle recognition device according to the invention of claim 8 is:
8. In addition to the same image input means and three-dimensional measuring means as in claim 7, each three-dimensional coordinate measured by said three-dimensional measuring means is projected on a virtual plane positioned perpendicular to the bottom surface of said observation area. Projection means, a projection model obtained by projecting the front of the specific vehicle type on the virtual plane, scans the virtual plane after the projection processing, performs matching processing, and performs matching processing on the model by the matching processing. When a corresponding projection image is extracted, the image processing apparatus includes a recognition unit that recognizes a spatial position corresponding to the extracted position as a head position of a vehicle of the specific vehicle type.

[0017]

According to the first and fifth aspects of the present invention, an image area corresponding to the front of a vehicle of a specific vehicle type is detected on an image obtained by the imaging means to determine whether a vehicle of the specific vehicle type exists. Since the determination is made, it becomes possible to easily and accurately detect the vehicle of the specific vehicle type by using the characteristics of the front surface of the vehicle body of the specific vehicle type.

According to the second and sixth aspects of the present invention, the model image representing the front of the vehicle of the specific vehicle type is scanned on the image obtained by the image pickup means and the matching process is carried out. The corresponding image area can be detected.

According to the third and seventh aspects of the present invention, the spatial position of the specific type of vehicle is easily and accurately detected by searching the three-dimensional measurement results for the measurement data corresponding to the front of the specific type of vehicle. can do.

According to the fourth and eighth aspects of the present invention, each three-dimensional coordinate measured in the observation area is projected on a virtual plane positioned perpendicular to the bottom surface of the observation area.
A matching process is performed using a projection model obtained by projecting the front of the vehicle of the specific vehicle type, and the vehicle of the specific vehicle type is detected. Therefore, a projection model of the front portion of the vehicle is used for the projection image on the virtual plane. It is possible to detect a specific type of vehicle only by the matching process.

[0021]

FIG. 1 shows an example of installation of a traffic flow measuring device according to an embodiment of the present invention. This traffic flow measuring device 41
Using the image obtained by imaging the road RD, the road RD
This is for measuring the number of vehicles and the moving speed of the vehicles above, and is provided with a camera 1 and a control device 2 on a column 45 installed near the road RD.

The camera 1 is supported at a predetermined height position with its optical axis obliquely downward and in the length direction of the road RD. The control device 2 sequentially captures image data from the camera 1, performs a process of detecting vehicles on the image and a process of measuring the number of vehicles, and transmits the processing results to a remote management center 4.
0 (shown in FIG. 15).

FIG. 2 shows an electrical configuration of the traffic flow measuring device 41. The control device 2 includes a CPU 5, a ROM 6,
An input image memory 9, a reference image memory 10, and a difference image memory 1 are transmitted to a control unit 4 comprising a RAM 7 via a bus 8.
1, a difference processing circuit 12, a large vehicle detection unit 13, a general vehicle detection unit 14, an output unit 15, and the like are connected.

In FIG. 1, reference numeral 3 denotes an A / D conversion circuit for digitally converting an analog image signal from the camera 1, and a digital gray scale image generated by the A / D conversion circuit 3. The data is taken into the input image memory 9 and held until image data for the next frame is input.

The reference image memory 10 stores, as a reference image for performing background subtraction processing on the input image, an image captured in a state where no vehicle exists on the road RD or obtained in a certain period in the past. The integration processing image of the plurality of input images thus obtained is stored. Difference processing circuit 12
Generates a difference image in which an image of an object other than a background such as a vehicle is extracted by performing difference processing on both pixel data for each set of pixels corresponding to the input image and the reference image.

FIG. 3A shows an example of an input image from the camera 1, in which a background image such as a road boundary or a nearby tree appears in addition to a vehicle image. Fig. 3 (2)
Is an image obtained by performing a difference process on the input image, and only an image of a vehicle to be detected appears. L in FIG. 3 (2) is a measurement line virtually set by the control unit 4 in order to recognize a passing vehicle.

Returning to FIG. 2, the CPU 5 stores the difference image in the difference image memory 11, and thereafter, using the large vehicle detection unit 13 and the general vehicle detection unit 14, the vehicle appearing on the difference image The vehicle type and position are detected. Further, the CPU 5 measures the moving speed of each vehicle on the road, the number of vehicles passing through the measurement line L, and the like using the detection result, and outputs the measurement result to the management center 40 via the output unit 15. Send to

The RAM 7 temporarily stores work data generated during the above-described series of processing, and also stores vehicle data obtained within a predetermined period as data for calculating the number of vehicles and the traveling speed. The detection results are stored in association with each other on the time axis. The ROM 6 stores a program indicating a processing procedure to be executed by the CPU 5 and fixed data such as a set position of the measurement line L.

The control device 2 of this embodiment focuses on the fact that the shape and size of the front windshield differ greatly between a large vehicle and a general vehicle. An image area corresponding to the windshield of a large vehicle is searched on the image.

Generally, a windshield of a passenger car has a trapezoidal shape spreading downward, while a large vehicle such as a truck or a bus has a large windshield of a rectangular shape.
Therefore, when both a large vehicle and a general vehicle appear on the image, the above-described differences in shape and size should also occur in the windshield of each vehicle type on the image.

A model image representing the characteristics of a typical windshield of a large vehicle is set in the large vehicle detection unit 13. By scanning this model image on the difference image and performing a matching process, Then, an image area corresponding to the windshield of a large vehicle is extracted. CPU5
Recognizes the extraction position of this image region as the leading position on the image of the large vehicle, and determines that the image region of a predetermined size based on this extraction position corresponds to the image of the large vehicle.

On the other hand, a model image representing the shape of the entire vehicle body of the general vehicle is set in the general vehicle detecting section 14. After the detection processing of the large vehicle, the model image of the general vehicle is set on the difference image. Then, by performing scanning and performing the same matching process, a general vehicle on the image is detected.

FIG. 4 shows a series of control procedures performed by the control device 2 for one frame of the image signal from the camera 1. In the figure, “ST” means each step. First, in step 1, the grayscale image data generated by the A / D conversion processing is stored in the image memory. In the next step 2, background difference processing using the reference image is performed, and a difference image including only an image of an object such as a vehicle is generated.

When the difference image is stored in the difference image memory 11, in the next step 3, the large vehicle detection unit 13
Performs scanning by setting the model image of the windshield on the difference image, and performing a correlation operation between the image region superimposed on the model image and the model image at each scanning position. When an image region r (shown in FIG. 3B) having a high correlation value with respect to the model image is extracted by the matching process, the CPU 5 determines the extracted position as a detection position of a large vehicle and The image area of a predetermined size before, after, and from the extraction position is determined as an image of a large vehicle, and each determination result is stored in the RAM 7.

On a two-dimensional image, an object located farther from the imaging position appears smaller in an upper position on the image. Therefore, in order to perform the matching processing with high accuracy, the y coordinate value of the scanning position is required. It is necessary to change the size of the model image according to the condition. For the same reason, it is desirable that the area of the image area recognized as the image of the large vehicle is also changed in accordance with the y-coordinate value of the set position. In the matching process, if the scanning region of the model is limited to a predetermined region including a road on the image, the processing time can be reduced.

When the detection process of the large vehicle is completed, the vehicle model is set on the difference image by the general vehicle detection unit 14, and the same correlation calculation process is performed to detect the general vehicle on the image. (Step 4). In this matching process as well, it is necessary to change the size of the model image according to the y-coordinate value of the scanning position, as described above. The image region previously recognized as the image of the large vehicle is excluded from the target of the matching processing.

When a large vehicle and a general vehicle on the image are detected in this way, the CPU 5 takes in the detected positions and stores them in the RAM 7 (step 5). RA
As described above, the vehicle detection result obtained within the predetermined period is accumulated in M7, and in the next step 6, CP
U5 recognizes the detected moving route of each vehicle by associating the detection result in the current input image with the past detection result. Further, in step 7, the moving speed is calculated for each vehicle using the detected positions of the vehicle for a predetermined number of times in the current and past times.

Next, the CPU 5 compares the current detection position of each vehicle and the detection position in the preceding stage with the set position of the measurement line L shown in FIG. 3 (2). If there is a vehicle whose measurement line is located between the detection positions, the vehicle is recognized as a passing vehicle, and the number of passing vehicles is measured (steps 8 and 9).

When the moving speed of each vehicle on the road and the number of passing vehicles are measured as described above, the results of the measurement are transmitted to the outside via the output unit 15 in the next step 10. , And the process ends.

For a large vehicle that is likely to be erroneously detected as described above, the image area of the windshield that can obtain common features in this vehicle type is extracted. Can be greatly improved. In addition, since a large vehicle occupies a large image area on an image, when a matching process is performed using a model of the entire vehicle body, the configuration of a correlation operation circuit becomes complicated and costs increase. By performing the matching processing using the glass model image, the correlation operation circuit can be simplified and efficient processing can be performed.

In this embodiment, after the input image is subjected to the difference processing, each model is scanned on the difference image to perform the matching processing based on the gray image data. After the extraction, the matching process may be performed on the edge image. In this case, each pixel data used for the matching process is binary data, so that the correlation operation can be easily performed.

FIG. 5 shows another example of installation of the traffic flow measuring device 41, in which vehicles on the road RD are recognized by three-dimensional recognition processing using two cameras 1a and 1b. Each camera 1a, 1b has an equal focal length,
The optical axes are paralleled by a support member (not shown) toward the direction of the road RD, and the imaging surfaces are arranged vertically with the respective imaging planes positioned on the same plane. The control device 2 controls each camera 1
After sequentially taking the images from a and 1b and performing a three-dimensional measurement process using the principle of triangulation, a projection process and a matching process described later are performed using the measurement results, and each vehicle is separated and detected. I do.

As shown in FIG. 6, the control device 2 of this embodiment includes an A / D conversion circuit 3 corresponding to each of the cameras 1a and 1b.
a, 3b, image memories 9a, 9b, reference image memory 10
a, 10b, edge image memories 16a, 16b, a feature extracting unit 17, a three-dimensional coordinate measuring unit 18, a projection processing unit 19, a large vehicle detecting unit 20, a general vehicle detecting unit 2.
1, each unit such as the output unit 15 is connected to the control unit 4 via the bus 22. The control unit 4 is the same as that in the first embodiment, and the CPU 5 controls the above-described units in series based on a control program stored in the ROM 6 to detect the spatial position of the vehicle on the road. The number of passing vehicles and the moving speed of the vehicles are measured using the detection results.

The image data from each of the cameras 1a and 1b is
After being digitally converted by the corresponding A / D conversion circuits 3a and 3b, they are stored in the input image memories 9a and 9b. The feature extraction unit 17 includes a difference processing circuit and an edge extraction circuit. The feature extraction unit 17 applies a reference image memory 10 to the input images stored in the input image memories 9a and 9b, respectively.
After performing the difference processing using the background images a and 10b, the edge extraction circuit performs edge extraction processing on the image after the difference processing, and extracts feature points representing the outline of the vehicle.

The three-dimensional coordinate measuring section 18 associates the extracted feature points between the images, and then applies the two-dimensional coordinates to the corresponding feature point sets according to the principle of triangulation to obtain these feature points. The three-dimensional coordinates of the object point in the space corresponding to the point are calculated.

The spatial coordinate system on which the three-dimensional coordinate calculation processing is based is, for example, as shown in FIG.
The origin O is defined by the foot of the perpendicular dropped from the focal point C of b to the ground,
The width direction of the road RD is the X axis, the height direction is the Y axis, and the length direction is the Z axis. In the figure, B denotes each camera 1a,
1B, F is the focal length, H is the installation height of the lower camera 1b (corresponding to the distance between points C and O), θ
Indicates depression angles of the cameras 1a and 1b, respectively.

Since the cameras 1a and 1b are arranged vertically with their optical axes parallel and their imaging planes G _U and _GL positioned on the same plane, the cameras 1a and 1b fall on a predetermined object point P in space. The imaging position on each input image is shifted only in the vertical direction. Therefore, the object point P on each input image
Assuming that the coordinates of the image points P _U and P _L are (x, y _U ) (x, y _L ), the three-dimensional coordinates (X, Y, Z) of the object point P are It is calculated by equation (3).

[0048]

(Equation 1)

[0049]

(Equation 2)

[0050]

(Equation 3)

When the three-dimensional coordinates are calculated for each set of feature points corresponding to each input image in this manner, the projection processing unit 1
9 projects these three-dimensional coordinates on a virtual vertical plane (YZ plane) along the length direction of the road RD.

FIG. 8 shows an example of the result of projection on the YZ plane. The points specified by the Y and Z coordinates of the object point indicated by each feature point are plotted on the YZ plane, so that the points on the road RD are plotted. A projection image showing the side shape of each vehicle is generated.

In the large vehicle detection section 20, a projected image of a windshield obtained when the front part of the body of a typical large vehicle is projected on the YZ plane is set as a projection model. The large-sized vehicle detection unit 20 detects the Y after the projection processing.
The projection model is set and scanned on the Z plane, a matching process between the projection image and the model at each scanning position is performed, and the leading position of the large vehicle is detected.

On the other hand, a side-surface model of the entire vehicle body of the general vehicle is set in the general-vehicle detecting unit 21. When the detection processing of the large vehicle is completed, the side-surface model is set on the YZ plane. The position of the general vehicle is detected by the same matching process.

FIG. 9 shows a processing procedure for one frame of image signal from each of the cameras 1a and 1b in the control device 2. First, in step 1, each camera 1a, 1
b is subjected to A / D conversion and stored in the corresponding image memory.
, A background difference process and an edge extraction process are performed, and feature points on each input image are extracted.

The three-dimensional coordinate measuring unit 18 associates the extracted feature points with feature points P _U and P _L representing the same object point P in space by performing a correlation operation between input images. Then, for each of the associated feature points P _U and P _L , the two-dimensional coordinates (x, y _L ) (x, y _U ) of each point are substituted into the above equations (1) to (3), and The three-dimensional coordinates of the point P are calculated (Step 3).

When the three-dimensional coordinates of the object point commonly appearing on each input image are calculated, the processing shifts to the processing by the projection processing unit 19, and the three-dimensional coordinates are projected on the YZ plane. A projection image representing the side shape of the vehicle is generated (Step 4). Then, in the next step 5, a matching process is performed by the large vehicle detection unit 20, and a large vehicle on the road is detected.

FIG. 10 shows an example in which a matching process using the projection model of the windshield (indicated by _MF in the figure) is performed on the projection result of FIG. This projection model M
_F is generated by cutting out a portion of a windshield from a projection model (model m shown by a chain line in the figure) of a front part of a vehicle body of a large vehicle, and is initially generated such that a left end point of the model is located on the Y axis. After the setting, scanning is performed sequentially along the Z axis.

The large-sized vehicle detection unit 20 outputs
For projection point located in the vicinity of the projection model M _F at that time, each calculates the distance to the model M _F, and further calculates the average of each distance, the projected image and the model M _F representative thereof of each projection points And the degree of difference. Degree of difference for each scan position is compared with a predetermined threshold value, when the degree of difference is below the threshold value, and the scan position Z _A as the minimum value is detected, CPU 5, the scanning position Z _A Is recognized as the head position of a large vehicle. The CPU 5 also determines that the start position Z
Each projection points included in the region of just to the rear of the point Z _B predetermined distance d than _A, recognized as a projection in configuration point of a large vehicle, to be excluded from the following matching process.

As is apparent from the example of FIG. 10, the windshield of a large vehicle is projected at a higher position on the YZ plane than the windshield of a small vehicle, and the size and inclination angle between the two are remarkable. because there are differences, the matching process by the projection model M _F, the head position of large vehicles can be accurately detected. Moreover, since it is not necessary to be subjected to matching processing each projected point to Z _B from the detected position Z _A, you can simplify the configuration of the arithmetic circuit, and may be processing speed faster.

In the actual processing, the size and height of the windshield of a large vehicle may vary slightly. However the inclination angle of the windshield large vehicles are the gentler than ordinary vehicle, by setting the combined projection model M _F at a height of a typical windshield, the windshield the projection model M _F Can be almost superimposed on a part of the projected image, and the matching processing can be performed without any trouble. Further, if performing the matching process while changing the height position of the projection model M _F within a predetermined range, it is possible to further improve the detection accuracy.

Returning to FIG. 9 again, when the processing of the large vehicle detecting section 20 is completed, the processing shifts to the processing of the general vehicle detecting section 21 and the above-described side shape model of the general vehicle is scanned on the YZ plane. Then, a similar matching process is performed (step 6). In this process, if a large vehicle is detected in step 5, each point recognized as a projection point of that vehicle is removed from the target of the current matching process. There is no possibility that a part of the image is erroneously detected as a general vehicle.

When the vehicle detection processing for each vehicle type is completed, each detection result is stored in the RAM 7 in the next step 7, and thereafter, in the same manner as in the first embodiment, each detected vehicle is detected. The speed and the number of passing vehicles are measured, and the measurement results are output (steps 8 to 12).

The measurement line used in the process of judging a passing vehicle in step 10 is set at a predetermined position on an actual road, and the current and previous detection positions of a certain vehicle are determined by the measurement line of this measurement line. When the vehicle is located across the set position, the vehicle is recognized as a vehicle that has passed the measurement line.

In the above embodiment, two cameras 1a, 1
Although the three-dimensional measurement processing using the image obtained by imaging the road by b is performed, the present invention is not limited to this, and the three-dimensional measurement processing may be performed by other means.

FIGS. 11 and 12 show the traffic flow measuring device 4.
1 shows another configuration. This apparatus extracts a feature indicating a three-dimensional shape of a vehicle using an invisible light beam such as a laser beam, and controls one camera 1 on a support 3 in the same manner as in the first embodiment. In addition to supporting the device 2, a light projecting device 25 is provided at a position before the support 3.

The light projecting device 25 irradiates a plurality of laser slit lights in parallel along the width direction of the road RD. The camera 1 has a filter adapted to the wavelength band of the laser slit light. Has been deployed.

The control unit 2 includes an A / D conversion circuit 3, an input image memory 9, a control unit 4 and an output unit 15, as well as a binarization circuit 26 and a binary image memory 27 as in the first embodiment. ,
It includes a reference image memory 28, a three-dimensional coordinate measuring unit 29, a projection processing unit 30, a large vehicle detecting unit 31, and a general vehicle detecting unit 32. As means for controlling the light intensity and light emission timing of the light emitting device 25, a light emitting control circuit 2
4 is added.

When the image data from the camera 1 is stored in the input image memory 9, the binarization circuit 26 binarizes the input image with a predetermined threshold value, and generates the slit light on the image. Extract the refraction pattern. This binary image is stored in the binary image memory 27 and is used for measurement processing by the three-dimensional coordinate measurement unit 29.

FIG. 13 shows an irradiation pattern of the slit light appearing on the binary image. If no vehicle exists in the slit light irradiation direction, the slit light travels straight along the width direction of the road, so that a parallel slit light irradiation pattern appears on the binary image. On the other hand, when a vehicle exists in the irradiation direction of the slit light, the slit light is reflected on the surface of the vehicle body, so that the vehicle (indicated by a chain line in the drawing) is located at the vehicle position on the binary image. A refraction pattern of the slit light that reflects the shape appears.

Returning to FIG. 12, the reference image memory 28 of this embodiment stores an irradiation pattern (hereinafter, referred to as a “reference pattern”) when the slit light is irradiated in a state where no vehicle exists on the road RD. The stored binary image is stored. After associating each constituent point of the refraction pattern of each slit light appearing on the binary image of the input image with the corresponding constituent point of the reference pattern, the three-dimensional coordinate measuring unit 29 calculates the distance between the corresponding points. The height data (Y coordinate) of each component point of the refraction pattern is calculated from the two-dimensional coordinates. X and Z coordinates corresponding to the respective constituent points of the reference pattern are set in the three-dimensional coordinate measuring unit 29. By applying these coordinate values to the respective constituent points of the refraction pattern, a slit on the vehicle body surface is obtained. The three-dimensional coordinates of each irradiation position of light can be derived.

As in the second embodiment, the projection processing unit 30 sets a virtual vertical plane (YZ plane) along the length direction of the road RD, and sets each three-dimensional coordinate calculated on this YZ plane. Is projected. The large vehicle detection unit 31 and the general vehicle detection unit 32 perform the same vehicle detection processing as described above on this projection image, and detect the leading position of the vehicle for each vehicle type.

FIG. 14 shows an example of a matching process for detecting a large vehicle. On the YZ plane, a vertical line image (shown by a solid line in the figure) corresponding to the length of the slit light applied to the side of the vehicle is generated by the projection processing of the refraction pattern of the slit light.

The large vehicle detecting section 31 connects the upper end points of these vertical line segment images to generate a contour pattern on the side of the vehicle (indicated by a dotted line in the figure), and then forms the same projection model M on the YZ plane as described above. By scanning _F , matching processing is performed on the contour shape pattern to detect the leading position of a large vehicle.

Similarly, the general vehicle detecting section 32 performs a matching process between the side shape model of the general vehicle and the contour shape pattern, and detects the head position of the general vehicle. Thereafter, as in the above-described embodiments, after the detection position of each vehicle is recognized by the CPU 5, measurement processing of the speed of the vehicle and the number of passing vehicles is performed.

In each of the above embodiments, a large vehicle is detected by extracting the characteristics of the windshield of a large vehicle. However, not only the windshield but also the characteristics of the entire front surface of the vehicle. You may make it extract. Similarly, for a general vehicle, the vehicle may be detected based on the features of the front glass and the front shape of the vehicle body.

In the second and third embodiments, the two-dimensional recognition process is performed after projecting the three-dimensional measurement result on a virtual vertical plane along the length direction of the road RD. The present invention is not limited to this, and the present invention can be applied to a case where stereoscopic recognition processing in a spatial coordinate system is performed using each measured three-dimensional coordinate.

As described above, the measurement data obtained by the traffic flow measuring device of any of the first to third embodiments is transmitted to a management center or the like, and the transmitted measurement data is It is used for various purposes, such as being edited as information and provided to a driver at a downstream position, or used for signal control processing.

FIG. 15 shows an example of a traffic information system, in which one of the above-mentioned traffic flow measuring devices 41 is installed at a plurality of points, and measurement data transmitted from each device is collected and edited by a management center 40. Then, it is output as traffic information at each point. In this case, each traffic flow measuring device 4
In either case, since the measurement data with high accuracy is transmitted by the above-described processing method, the management center 40 can obtain more accurate traffic flow data than before, and the processing accuracy of the entire system can be greatly improved. it can.

Here, an apparatus for detecting a vehicle on a road and performing a traffic flow measurement has been described. However, the present invention is not limited to this, and is also applicable to an apparatus for observing a vehicle parked in a parking lot or the like. can do.

[0081]

As described above, the present invention is applicable to a case where a vehicle of a specific vehicle type is detected by two-dimensional image processing or three-dimensional measurement processing using an image obtained by imaging a predetermined observation area by an imaging means. Furthermore, since the detection process is performed with the feature representing the front of the vehicle of this type of vehicle, the vehicle of the specific type of vehicle can be easily and accurately detected, and the highly accurate vehicle recognition process can be performed.

Further, according to the fourth and eighth aspects of the present invention, each three-dimensional coordinate measured in the observation area is projected on a virtual plane positioned perpendicular to the bottom surface of the observation area, and then the projection image is specified. Since the matching process is performed using a projection model obtained by projecting the front of the vehicle of the vehicle type and the vehicle of the specific vehicle type is detected, the projection model of the front portion of the vehicle is used for the projection image on the virtual plane. A specific type of vehicle can be detected only by the matching process, and the processing time can be significantly reduced.

[Brief description of the drawings]

FIG. 1 is a perspective view showing an example of installation of a traffic flow measuring device to which the present invention is applied.

FIG. 2 is a block diagram showing a configuration of the traffic flow measuring device of FIG.

FIG. 3 is an explanatory diagram showing an input image and a difference image after a background difference process.

FIG. 4 is a flowchart illustrating a processing procedure in the control device.

FIG. 5 is a perspective view showing another example of the traffic flow measuring device.

FIG. 6 is a block diagram showing a configuration of the traffic flow measuring device of FIG.

FIG. 7 is an explanatory diagram showing an example of setting a spatial coordinate system for a three-dimensional measurement process.

FIG. 8 is an explanatory diagram showing a state where three-dimensional coordinates are projected on a YZ plane.

FIG. 9 is a flowchart illustrating a processing procedure in the control device of FIG. 6;

FIG. 10 is an explanatory diagram showing a detection processing method for a large vehicle.

FIG. 11 is a perspective view showing a third example of the traffic flow measuring device.

FIG. 12 is a block diagram showing a configuration of the traffic flow measuring device of FIG.

FIG. 13 is an explanatory diagram showing a refraction pattern of laser slit light on an image.

14 is an explanatory diagram showing a state in which a detection process of a large vehicle is performed using a projection result of three-dimensional coordinates obtained from the refraction pattern of FIG.

FIG. 15 is an explanatory diagram showing an example of a traffic information system using a traffic flow measuring device.

[Explanation of symbols]

 1, 1a, 1b Camera 2 Control device 5 CPU 13, 20, 31 Large vehicle detecting unit 18, 29 Three-dimensional coordinate measuring unit 19, 30 Projection processing unit

Claims (8)

[Claims] 1. A method for recognizing a vehicle located in an observation area using an image obtained by imaging a predetermined observation area by an imaging means, the image corresponding to a front of a vehicle of a specific vehicle type on the image. Search for the corresponding image area, and when the corresponding image area is detected,
A vehicle recognition method comprising recognizing that the vehicle of the specific vehicle type exists at a spatial position corresponding to the detected position. 2. A method for recognizing a vehicle located in an observation area using an image obtained by imaging a predetermined observation area by an imaging unit, wherein a model image representing a front surface of a specific vehicle type is displayed on the image. And a matching process is performed by scanning the model image. When an image region corresponding to the model image is extracted, it is recognized that a vehicle of the specific vehicle type exists at a spatial position corresponding to the extracted position. Method. 3. A three-dimensional measurement process in the observation area is performed by using an image obtained by imaging a predetermined observation area by at least one imaging unit, and a result of the three-dimensional measurement in the observation area is obtained by using the measurement result. In the method for recognizing a vehicle located at a specific vehicle type, a measurement result corresponding to the front of a vehicle of a specific vehicle type is searched from the three-dimensional measurement results, and when the corresponding measurement result is detected, the specific vehicle type is detected at the detected position. A vehicle recognition method characterized by recognizing that a vehicle exists. 4. A three-dimensional measurement process in the observation region is performed by using an image obtained by imaging a predetermined observation region by at least one imaging unit, and a result of the three-dimensional measurement in the observation region is calculated using the measurement result. In the method of recognizing a vehicle located at, the three-dimensional coordinates measured by the three-dimensional measurement processing are:
After projecting on a virtual plane located perpendicular to the bottom surface of the observation area, a projection model of the front of a specific vehicle type is scanned on this virtual plane, and when a projection image corresponding to this model is extracted, the extracted position is extracted. A vehicle recognition method comprising: recognizing that a vehicle of the specific vehicle type exists at a spatial position corresponding to (i). 5. An image input means for inputting an image obtained by imaging a predetermined observation area by an image pickup means, and an image area corresponding to a front of a vehicle of a specific vehicle type on the image input by the image input means. And a recognizing means for recognizing, when the image area is detected by the searching means, an image of the vehicle of the specific vehicle type at the detected position. 6. An image input means for inputting an image obtained by imaging a predetermined observation area by an image pickup means, and scanning a model image representing the front of a specific vehicle type on the image input by the image input means. A matching processing means for performing a matching process; and recognizing that when an image region corresponding to the model image is extracted by the matching process, an image of the vehicle of the specific vehicle type is present at an extraction position of the image region. And a vehicle recognition device. 7. An image input unit for inputting an image obtained by imaging a predetermined observation region by at least one imaging unit, and using the image input by the image input unit to obtain a 3D image in the observation region. Three-dimensional measurement means for performing a dimension measurement process; search means for searching measurement data corresponding to the front of a vehicle of a specific vehicle type from the three-dimensional measurement results; and when the measurement data is detected by the search means ,
A vehicle recognizing device comprising: recognizing means for recognizing the detected position as a head position of a vehicle of the specific vehicle type. 8. An image input unit for inputting an image obtained by imaging by at least one imaging unit, and a three-dimensional measurement process in the observation region is performed using the image input by the image input unit. Three-dimensional measuring means, and three-dimensional coordinates measured by the three-dimensional measuring means,
The projection means for projecting on a virtual plane positioned perpendicular to the bottom surface of the observation area, and a projection model obtained by projecting the front of a specific vehicle type on the virtual plane, scans the virtual plane after the projection processing, and performs matching. A matching means for performing processing; and a recognition means for recognizing, when a projection image corresponding to the model is extracted by the matching processing, a spatial position corresponding to the extracted position as a head position of a vehicle of the specific vehicle type. A vehicle recognition device comprising the same.