JPH11353581A - Method and device for discriminating vehicle kind in the daytime - Google Patents

Abstract

PROBLEM TO BE SOLVED: To discriminate the kind of a vehicle with a high precision and to discriminate more vehicle kind. SOLUTION: The picture of a television camera is differentiated and is stored in a differential picture storage part 13. A road surface eliminating means 12a projects the differential picture and collects the luminance of the road surface to eliminate the road surface. A head detection means 12b detects the head position of a vehicle. A vehicle width detection means 12b detects the vehicle width of the vehicle. A rear detection means 12d detects the rear position of the vehicle. A silhouette segmenting means 15a projects a partial picture to segment the silhouette of the vehicle. A part discrimination means 15b discriminates each part of the vehicle. A vehicle kind discrimination part 17 refers to features data of vehicles stored in a feature data memory 20 to discriminate the kind of the vehicle as a large-sized or small-sized vehicle and as a large-sized truck, a large-sized passenger car, a small-sized truck, or a small- sized passenger car based on the silhouette of the vehicle and the differential picture.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a vehicle type discriminating apparatus for discriminating a vehicle type at a measuring point, and more particularly to a vehicle type discriminating apparatus and method during daytime.

[0002]

2. Description of the Related Art A vehicle detecting device for monitoring the traffic flow of a vehicle running on a road can monitor the congestion status, the number of vehicles passing therethrough, the vehicle speed, and the like, and is used for road management, planning, environmental prediction, and the like. The vehicle detection device performs a predetermined image processing on a vehicle imaged by a camera and performs each of the above measurements.

For example, in a first invention disclosed in Japanese Patent Application Laid-Open No. 5-290293, a predetermined area on a road is photographed by a camera device, an edge portion in this area is taken out, and a leading position of the vehicle is detected. It is. Also, Japanese Patent Application Laid-Open No. 6-33
The second invention disclosed in Japanese Patent No. 7998 is a configuration for executing edge detection processing and movement information detection processing on an image captured by a vehicle monitoring camera. This is a highly accurate moving object measurement of a vehicle.

[0004]

However, the first invention of the prior art employs a configuration in which edge processing is performed only on a predetermined area of an image to detect the head, and the vehicle type cannot be accurately determined only from the head. . Further, if there is a shadow component of a vehicle or the like in the image, the position of the head of the vehicle cannot be accurately detected due to the influence of the shadow. Further, in the second invention, since the moving object of the vehicle is measured using the time-series data, it cannot cope with a sudden change in the surrounding environment (for example, a sunshine state). Also, the vehicle type cannot be determined. In traffic flow monitoring, a means for accurately specifying the type of vehicle is desired.

SUMMARY OF THE INVENTION The present invention has been made to solve the above problems, and has as its object to provide a daytime vehicle type discriminating apparatus and method which can discriminate a vehicle type with high discrimination accuracy and can discriminate more vehicle types. I have.

[0006]

In order to achieve the above object, a vehicle type discriminating apparatus according to the present invention comprises:
A vehicle type discriminating device for discriminating a vehicle type of a road traveling vehicle during daytime, comprising: a camera (1) for capturing an image of a traveling vehicle; and removing a road surface and detecting a vehicle head position based on an image captured by the camera. A processing unit (2) for determining a vehicle type based on a vehicle width and a vehicle silhouette at the vehicle head position.

According to a second aspect of the present invention, the processing section (2) includes a road surface removing means (12a) for differentiating the image and removing a road surface from the image based on the differentiated image. Is also good.

The road surface removing means (12a) may project the differentiated image in the X-axis and Y-axis directions to determine and remove the road surface.

Further, as set forth in claim 4, the road surface removing means (12a) projects the differentiated image in the X-axis and Y-axis directions to obtain an average value of projection values of each axis, and A configuration in which a range having a differential value lower than the average value on the axis or the Y axis may be determined as a road surface and removed.

Further, the road surface removing means (12a) projects the differentiated image in the X-axis and Y-axis directions to obtain an average value of the projection values of each axis, and Luminance values in a range having a differential value lower than the average value on the Y axis or the Y axis are collected, sorted in the order of most abundance, and the luminance values in a predetermined range centered on the luminance values from the top to a certain rank or a certain value. May be determined to be a road surface.

According to a sixth aspect of the present invention, the processing unit (2) differentiates the image and removes a road surface from the image based on the differentiated image, thereby providing a temporary vehicle width and vehicle presence. A position candidate is obtained, an image of the existence position candidate is projected on the Y-axis, Y-axis lines having a projection value equal to or more than a predetermined value are collected to form a temporary head candidate group, and a Y-axis line of the temporary head candidate group is collected. Of these, a head detection unit (12b) for determining the Y-axis line at the lowermost position of the image as the temporary head position of the vehicle may be provided.

According to a seventh aspect of the present invention, the head detecting means (12b) sorts a predetermined range of luminance projection values based on the Y-axis line obtained as the tentative head candidate. In the order of darkness up to the predetermined order,
If a Y-axis line located at the lower part of the image from the temporary head candidate is detected, the detected Y-axis line may be determined as the head position of the vehicle.

The processing section (2) differentiates the image of the vehicle head position based on the detected vehicle head position and performs X-axis control within a certain range based on the vehicle head position. The vehicle may be configured to include a vehicle width detection unit (12c) that determines the vehicle width by projecting and obtaining a projection value equal to or greater than a certain value from the left and right.

According to a ninth aspect of the present invention, the processing section (2) has a length within a certain range based on a vehicle length set in correspondence with the detected vehicle width. May be differentiated onto the Y-axis to obtain a projection value equal to or greater than a certain value (lower limit average value). .

[0015] According to a tenth aspect of the present invention, the rear detecting means (12d) is configured to control the Y position at the head position of the vehicle.
When the deviation in the X-axis direction between the center of gravity of the vehicle width partial line of the axis and the center of gravity of the differential value for each line of the Y axis is within a predetermined range, the determined rear position is the rear position of the vehicle. On the other hand, if it is determined that there is a vehicle, and the vehicle exceeds the predetermined range, it is determined that the vehicle is a following vehicle, and the next candidate satisfying the condition is not detected as the vehicle rear, but may be detected as the vehicle rear.

A vehicle type discriminating apparatus according to an eleventh aspect of the present invention is a vehicle type discriminating apparatus for discriminating a vehicle type of a road traveling vehicle during daytime, wherein a camera (1) for imaging the traveling vehicle and an image captured by the camera. A vehicle position determining unit (12) for determining a head position, a vehicle width, and a rear position of the vehicle in the image based on the image, and determining an existing position of each part of the vehicle based on the determination information of the vehicle position determining unit. And a vehicle type determining unit (17) for determining a vehicle type based on information on each part of the vehicle determined by the vehicle region determining unit.
And characterized in that:

According to a twelfth aspect of the present invention, the vehicle position judging section (12) differentiates the image based on an image taken by the camera, and removes a road surface from the image based on the differentiated image. Road surface removing means (12a) for performing a differentiation on the image, removing a road surface from the image based on the differentiated image to obtain a tentative vehicle width and a vehicle existing position candidate, and obtaining the image of the existing position candidate. Projected to the Y-axis, Y-axis lines having a projection value equal to or greater than a predetermined value are collected and set as temporary leading candidates, and the Y-axis line at the lowermost position of the image among the Y-axis lines of the temporary leading candidate group is A head detecting means (12b) for determining a temporary head position; a means for determining a head position based on the temporary head position and a luminance projection value in the vicinity thereof; and a vehicle head position based on the detected vehicle head position. Differentiate the image of the head of the vehicle The vehicle width detecting means (12c) for obtaining the vehicle width by projecting the X axis in a certain range based on the position and obtaining a projection value equal to or more than a certain value from the left and right, and is set corresponding to the detected vehicle width. Rear detection means (12d) for determining the rear position of the vehicle based on the vehicle length; and a vehicle part for determining each part of the vehicle based on the information of the head position, the vehicle width, and the rear position of the vehicle after the road surface is removed. Judgment unit (1
5).

According to a thirteenth aspect of the present invention, the vehicle part judging section (15) includes a left end of a left end edge of a rear position of the vehicle and a right end of a right end edge detected based on a differentiated image obtained by differentiating the image. It is also possible to provide a silhouette cutting means (15a) for judging edges continuous in the respective Y-axis coordinates as the left end and the right end of the vehicle and cutting out the silhouette of the vehicle until reaching the head position of the vehicle.

According to a fourteenth aspect of the present invention, the silhouette extracting means (15a) forms a line continuous at a certain or more differential value in the X-axis direction based on a differential value obtained by projecting the differential image in the Y-axis direction. It is judged as an effective edge, and an effective edge that extends at least to the lower part of the image with respect to the left end of the leftmost edge and the right end of the rightmost edge among the effective edges at the rear of the vehicle is judged as the left end and the right end of the vehicle A configuration may be employed in which the silhouette of the vehicle is cut out until the head position is reached.

According to a fifteenth aspect of the present invention, the vehicle part judging section (15) sets the vehicle into a head portion (height of a hood) based on a ratio of a vehicle width to a vehicle length of the cut-out silhouette. A part determining means (1) for determining each part of a bonnet part, a windshield part, and a rear part after a roof.
5b).

According to a sixteenth aspect of the present invention, in addition to the part determination based on the silhouette, the part determination means (15b) projects a differential image obtained by differentiating an image captured by the camera in the Y-axis direction. Each part of the vehicle may be determined based on the ratio between the vehicle width and the vehicle length based on the edge obtained as described above.

According to a seventeenth aspect of the present invention, a feature data memory (20) in which feature data for each vehicle type is stored in advance.
The vehicle type determination unit (17) may be configured to determine the vehicle type by referring to the determination information of the vehicle part determination unit with reference to the characteristic data in the characteristic data memory.

Further, the vehicle type determining section (17) stores the determination information at the time of the vehicle type determination in the characteristic data memory (20) every time the vehicle type is determined. Is also good.

According to a nineteenth aspect of the present invention, the vehicle type determining section (17) is based on the vehicle width and / or the area of each part at the head position of the vehicle determined by the vehicle part determining section (15). Alternatively, it may be configured to determine whether the vehicle is a large vehicle or a small vehicle.

According to a twentieth aspect of the present invention, the vehicle type judging section (17) determines whether the vehicle type is a large vehicle when the result of the judgment is a large vehicle.
It may be configured to determine whether the vehicle is a large cargo vehicle or a large passenger vehicle based on the area ratio between the windshield and the hood.

According to a twenty-first aspect, when the result of the determination is that the vehicle is a small vehicle,
It may be configured to determine whether the vehicle is a small cargo vehicle or a small passenger vehicle based on the ratio of the vehicle width to the vehicle length and / or the height of the hood.

According to a twenty-second aspect of the present invention, there is provided a vehicle type determining method for determining a vehicle type of a road traveling vehicle during daytime, wherein the traveling vehicle is imaged by a camera, and the vehicle is imaged by the camera. After the differentiated image is differentiated, the differentiated image is projected in the X-axis and Y-axis directions to determine and remove the road surface, and then the vehicle type is determined based on the vehicle width at the head position of the vehicle and the silhouette of the vehicle.

Further, as set forth in claim 23, the differentiated image is projected in the X-axis and Y-axis directions to obtain an average value of the projection values in each axis, and the average value in the X-axis or Y-axis is calculated from the average value. A configuration in which a range having a low differential value is determined as a road surface may be adopted.

Further, as set forth in claim 24, the differentiated image is projected in the X-axis and Y-axis directions to determine the average value of the projection values of each axis, and the average value is calculated from the average value in the X-axis or Y-axis. A configuration may be adopted in which luminance values in a range having a low differential value are collected, and a pixel having a luminance value in a predetermined range up to a certain order or with a certain value as the center in the descending order of the luminance value is determined as a road surface.

According to a twenty-fifth aspect of the present invention, a temporary vehicle width and a candidate vehicle position are obtained by removing a road surface from the image based on the differentiated image. The Y axis line having a projection value equal to or greater than a predetermined value is collected as a temporary head candidate, and the Y axis line at the lowermost position of the image among the Y axis lines of the temporary head candidate is set as the head position of the vehicle. May be determined.

Further, the Y-axis line luminance projection values obtained as the tentative head candidate are sorted, and the Y-axis line luminance projection values are sorted in the order from the darkest luminance value to a predetermined order.
When a Y-axis line located within a certain range from the temporary head position below the image from the temporary head candidate is detected, the detected Y-axis line is determined to be the head position of the vehicle. Is also good.

According to a twenty-seventh aspect of the present invention, the image is differentiated and the X-axis is projected in a certain range based on the vehicle head position, and a projection value equal to or more than a certain value (lower limit average value) is obtained from the left and right. A configuration in which the vehicle width at the head position is obtained may be adopted.

According to a twenty-eighth aspect of the present invention, a rear portion of the vehicle is determined based on a vehicle length set in correspondence with the detected vehicle width, and a vehicle width portion of the Y axis at a leading position of the vehicle is determined. When the deviation in the X-axis direction between the barycenter position of the line and the barycenter position of the differential value of each line on the Y axis is within a predetermined range, it is determined that the determined rear position is the rear position of the vehicle. Alternatively, a configuration may be adopted in which, when the predetermined range is exceeded, the subsequent candidate satisfying the condition is not detected as the rear of the vehicle but is detected as the rear of the vehicle.

According to a twenty-ninth aspect, a line that is continuous with a certain differential value or more in the X-axis direction is determined as an effective edge based on a differential value obtained by projecting the differential image in the Y-axis direction. The effective edge, which extends at least to the lower part of the image from the left edge of the left edge and the right edge of the right edge of the effective edges to the lower part of the image, is determined to be the left edge and the right edge of the vehicle, and the silhouette of the vehicle until the leading edge of the vehicle And then, based on the ratio of the car width to the car length of the cut out silhouette, the vehicle is distinguished from its head portion (hood height), hood portion, windshield portion, and rear portion after the roof. It is good also as a structure which performs.

According to a thirtieth aspect of the present invention, whether the vehicle is a large vehicle or a small vehicle is determined based on the determined vehicle width at the head position and data of each part.
It may be configured to determine whether the vehicle is a large cargo vehicle or a large passenger vehicle, and whether the vehicle is a small cargo vehicle or a small passenger vehicle.

According to the above configuration, an image of a traffic flow of a moving body approaching from a distance is captured by the television camera 1, and the vehicle type is determined based on the captured image. The processing unit 2 differentiates the image of the camera 1, removes the road surface based on the differentiated image, determines the head position, the vehicle width, and the rear position of the vehicle, and obtains information on each part of the vehicle. The vehicle type is determined based on information on the vehicle width and each part. In the vehicle type determination, large and small vehicles are distinguished, and large cargo, large passenger, small cargo, and small passenger are distinguished.

[0037]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Next, an embodiment of the daytime vehicle type discriminating apparatus according to the present invention will be described. FIG. 1 is a schematic diagram showing a road as an object for which the vehicle type discrimination device of the present invention discriminates a vehicle type. This vehicle type discrimination device can be applied to a vehicle detection device that monitors the traffic flow of a vehicle passing on a road. The present invention is intended to determine the vehicle type based on the vehicle width and the vehicle silhouette.

The vehicle travels on the road C in the direction of the arrow. The ITV camera 1 is provided at a predetermined height position (for example, 6 to 8 m on the road) on the road C, and captures an image of a predetermined distance along the road. The ITV camera 1 is fixed to a support P or the like installed on a road C.

FIG. 2 shows a road C captured by the ITV camera 1.
FIG. A short distance portion of the road C is located at the “ground” position of the image, and a long distance portion of the road C is located at the “heaven” position of the image. The device determines the type of vehicle traveling on each of the two lanes C1 and C2 in the predetermined range shown in FIG. Here, the ITV camera 1 is disposed above the center of the two lanes C1 and C2. As shown in the figure, the right side of the vehicle D1 running on the lane C1 is obtained at the same time as viewed from the ITV camera 1 side, and the left side of the vehicle D2 running on the lane C2 is also obtained at the same time as shown in the figure.

FIG. 3 is a block diagram showing the overall configuration of the present apparatus. The device is configured by computer hardware such as a personal computer and software of a vehicle determination program stored in a storage unit. The hardware executes the vehicle determination program to perform a vehicle determination described later.

As shown in the figure, the apparatus comprises one ITV camera 1 and a general-purpose computer PC such as a personal computer. The computer PC includes a processing unit 2 such as a CPU,
It has an image memory 3 such as an AM and a ROM, and a peripheral portion (not shown) such as an I / F. A display unit 5 such as a CRT is connected to the computer PC, and an image of the ITV camera 1, an image during the vehicle type determination process, and other operation and setting contents of the computer are displayed on the screen. Computer P
An operation unit 6 such as a keyboard is connected to C.
Various operations and settings are input using.

The image memory 3 is composed of an image memory
The video signal of the TV camera 1 is captured at the timing of vehicle detection. The present invention is not limited to this, and a configuration may be adopted in which a plurality of images are held at predetermined intervals and then an image at the time of vehicle detection is extracted.
Here, the image of the still image is, for example, 6 X and Y coordinates.
It is 40 × 480 pixels (images are taken about 6 m on the left and right and about 20 m on the front and back). This image is divided (640 × 240) into two lanes as a first (second) field image. For each lane, an image of 200 × 150 pixels is cut out from a 640 × 240 pixel within a designated range (substantially the center position) and used for vehicle type discrimination. In the case of a small vehicle, the image of the cut-out range fits in the entire vehicle.
In the case of a large vehicle, the driver's seat portion and about 5 m behind it fall within the range.

FIG. 4 is a block diagram showing the internal configuration of the processing unit 2. This processing unit 2 determines the type of vehicle in each lane,
In addition, the total number of passing vehicles of each vehicle type is output.

The image of the ITV camera 1 is input to an image input unit 10
Is input to This image has, for example, 8-bit light and shade gradations (256 gradations from 0 to 255). Image input unit 10
Stores the image at the time of vehicle passage detection in the image memory 3 as a digital image still image (referred to as an original image). The vehicle detection is performed, for example, on the basis of the luminance values (for example, 0 to 2) of a plurality of pixels in a predetermined range (the lower half of the lanes C1 and C2) of the image.
A time change of any of the 256 gray levels between 55 is detected, and if the change is rapid, it is determined that the vehicle is passing. The configuration of the vehicle detection is generally known.
Not limited to such image processing, a configuration may be adopted in which ultrasonic waves or infrared rays are irradiated onto the lane, and vehicle passing is detected by the reflected light. However, in this case, a configuration of a detector such as an ultrasonic generator and a receiver is required.

The image input section 10 has a function of determining day or night based on an image. The day / night determination is performed by measuring the brightness value of a plurality of pixels set in advance from among all the pixels constituting the image in the image memory 3 and comparing the measured brightness value with a predetermined threshold value to determine whether the day or night. I do. And
If the result of the determination is daytime, the subsequent sections are activated to execute vehicle type determination. If the result of the determination is night, the various components (not shown) are operated to execute the vehicle type determination at night.

Before performing the vehicle type discriminating process, the filter unit 11 performs a filtering process on the original image to extract only the vehicles in the target lane.

The vehicle position determining section 12 determines the position where the vehicle exists based on the original image after the filtering process.
The vehicle position determining unit 12 includes: 1) a road surface removing unit 12a for removing a road surface portion; 2) a head detecting unit 12b for detecting a vehicle leading portion; 3) a vehicle width detecting unit 12c for detecting a vehicle width of the vehicle leading portion; 4) Rear detection means 12 for detecting the rear of the vehicle
d.

1) Removal of Road Surface In order to determine the type of vehicle, preprocessing for extracting only the vehicle from the image as much as possible leads to an improvement in accuracy. Therefore, it is essential to remove the road surface with high accuracy. Since the target image is a vehicle under natural light, a method not affected by disturbance light is used. The edge component of the image is compared with the luminance component, and it can be said that it is hardly affected by the level of light at that time. Therefore, the road surface portion is detected using the edge component. For this reason, the road surface removing means 12a performs a differentiation process on the original image, and stores the differentiated image in the differential image storage unit 13. Further, the differential value is projected on the X-axis and the Y-axis, and the average of the X-axis and the Y-axis projected values and the respective lower thresholds are used. Judge and collect the brightness. Then, a pixel having this luminance is detected by scanning, and a process of removing a pixel having a value determined to be road surface luminance is performed.

2) Detection of Vehicle Head The head of the vehicle has an edge component, and the vehicle head often has a shadow of the vehicle body in that part. The head detector 12b first assumes that a shadow of the vehicle body exists as the vehicle head position, and obtains the vehicle head position using the edge and the shadow.

3) Vehicle width detection The vehicle width detection means 12c calculates the vehicle width of the vehicle head from the vehicle head position detected by the head detection means 12b. The vehicle width uses only the X-axis projection value of the differential value.

4) Detection of the rear part of a vehicle Since a large vehicle has a longer vehicle length than a small vehicle, the rear detecting means 12d uses a large and small vehicle based on the vehicle width information obtained by the vehicle width detecting means 12c. The vehicle type is temporarily determined. Based on the determination, the rear position of the vehicle is estimated.

The vehicle part determining section 15 determines the characteristics of the vehicle in a state where only the vehicle is cut out from the image and generates data for determining the type of the vehicle. Silhouette cutting means 15
“a” detects the silhouette (outer edge line) of the vehicle based on the head position, vehicle width, and rear position of the vehicle determined by the vehicle position determination unit 12. The part determining means 15b determines each part of the vehicle (light part, hood part, windshield part, roof, and the following parts) based on the vehicle width, vehicle length data, and data at the time of silhouette detection.

The vehicle type determining unit 17 determines the vehicle width and the vehicle length, the data of the vehicle part determined by the vehicle part determining unit 15, and the characteristic information of the vehicle type stored in the characteristic data memory 20. The vehicle is classified into a large vehicle and a small vehicle. Heavy vehicles are further classified into large passenger and large cargo. Small vehicles are further classified as small passenger and small cargo. The number counting unit 30 counts the determined vehicles for each vehicle type. The vehicle type determined by the vehicle type determination unit 17 and the counting result of the number counting unit 30 are displayed on the display unit 5. Further, it can be output to other processing means via a printer or a line (not shown).

Next, a series of flow of the vehicle type determination processing in the above configuration will be described with reference to a flowchart. FIG.
It is a flowchart which shows the vehicle type determination process which an apparatus performs. The original image captured by the ITV camera 1 (see FIG. 6)
The images are continuously input to the image input unit 10. Image input unit 10
Stores the input original image in the image memory 3 as a still image. Although the image in FIG. 6 has been subjected to predetermined image processing (error diffusion processing) for convenience, it is actually a black-and-white image in which each pixel has a predetermined gradation (for example, 256 gradations).

The filter section 11 filters the original image in the image memory 3 (SP2). First, a large noise (a car in an adjacent lane, etc.) is removed. The area required for the vehicle type determination is the entire vehicle in the case of a small vehicle, and the driver's seat portion and about 5 m behind the driver's seat in the case of a large vehicle, so other portions are removed. Since the data on the white line for dividing a plurality of lanes (lanes) has little effect on the vehicle type determination, the white line is removed by mask processing.

Next, the vehicle position determining unit 12 determines the position where the vehicle is present based on the still image (referred to as the original image) captured by the image input unit 10 (SP3). The method of processing with a single still image has the advantage that it is more resistant to disturbance and can obtain accuracy independent of the speed of the vehicle, as compared with the method of determining the vehicle position based on the time difference.

First, the road surface portion is removed by the road surface removing means 12a. FIG. 7 is a flowchart showing the road surface removal processing. The road surface removal process includes a process of selecting a vehicle existing portion (SA) and a process of collecting road surface luminance from a road surface portion (S).
B), determination of road surface luminance (SC), and road surface portion removal processing (SD).

In the vehicle existence section selection process (SA), first, a differential image (see FIG. 8) obtained by differentiating the original image is stored in the differential image storage section 13. The differential image in FIG. 8 has been subjected to black-and-white inversion processing and error diffusion processing for convenience. Actually, each pixel is a black-and-white image having a predetermined gradation (for example, 256 gradations). Then, the road surface removing unit 12a projects the differential image on the X axis and the Y axis (SA1).

Next, an average value of the projection values is obtained (SA
2). Next, a section that is equal to or more than the average value with respect to the X axis is obtained (S
A3). In addition, a section of the Y axis that is equal to or more than the average value is obtained (SA4). Then, an area (high edge portion) included in any of the X and Y axis sections is determined as a vehicle existing area (SA5). FIG. 9 is a determination image of the vehicle presence. In the image shown in FIG. 2A, the vehicle portion is displayed in reverse video by projecting the image on the X axis and the Y axis. The image in FIG. 9 has been subjected to black-and-white inversion processing and error diffusion processing for convenience. Actually, each pixel is a black-and-white image having a predetermined gradation (for example, 256 gradations).

In the road surface luminance collecting process (SB), luminance data outside the vehicle existing area is collected as luminance data in pixel units (SB1). That is, if the projection value is equal to or smaller than the average value (low edge portion), the road surface is determined, and the luminance of the road surface portion is collected. In FIG. 9B, a portion of FIG. 9A that is determined to be a road surface is indicated by oblique lines. At this time, the vehicle is prevented from entering the low edge portion. This is because if the road surface enters the high edge portion and the vehicle does not enter the low edge portion, the road surface luminance can be collected at the low edge portion.

Next, a road surface luminance determination process is executed (S
C). When the brightness of the road surface portion is known, the road surface portion can be determined. First, the luminance data is converted into the luminance abundance (number of pixels).
(SC1). Next, for example, the luminance of the abundance in the top ten places, starting with the luminance having the largest abundance (number of pixels), is set as the temporary road surface luminance (SC2). Also,
With respect to each of the top ten road surface luminances, the luminance up to a certain width (for example, ± 10th) is determined as the road surface luminance (SC3). In the road surface portion removal processing (SD), pixels corresponding to the road surface luminance are removed from the original image stored in the image memory 3. FIG. 10 is an image after road surface removal from the original image. In the figure, the road surface portion is removed in white.

It can be expected that a silhouette portion of a car is detected in a portion where the X-axis and Y-axis projection values are high. Even if the brightness of the vehicle is almost the same as the road surface brightness, the threshold used to detect the road surface portion is determined based on the projection value of the edge, so the vehicle portion always remains. There is no significant effect on subsequent processing. With this method, if the area of the shadow is large, the shadow can be removed. If the shadow is small, it cannot be removed, but it has less effect on vehicle type discrimination than a large shadow. After the road surface portion removal processing using the road surface luminance, the remaining noise is removed from a predetermined range on the left, right, and bottom of the image (other than the position where the vehicle exists). FIG. 11 is a diagram illustrating an original image after noise removal. In FIGS. 10 and 11, the vehicle portion has multiple gradations as in the original image. It is to be noted that the error diffusion processing is illustrated for convenience.

Next, the head position detecting means 12b determines the head position of the vehicle. FIG. 12 is a flowchart showing the process of determining the head position of the vehicle. The existence area of the vehicle is determined to some extent by the road surface removal processing. The leading portion of the vehicle has an edge component, and the leading portion of the vehicle often has a shadow of the vehicle body at that portion. The head detector 12b first assumes that a shadow of the vehicle body exists as the vehicle head position, and obtains the vehicle head position using the edge and the shadow. In the method using this shadow, a precise vehicle head position may not be obtained in some cases. However, a shape error of the vehicle, a vehicle width, and a vehicle length may be determined as long as the error is allowable. In addition, in the method of obtaining the vehicle head position using only the edge component, a method using a shadow is used since it is recognized that the head position is erroneously recognized when the edge component is weak, which has a serious effect on subsequent processing. Prioritize. When the shadow cannot be recognized, the head position of the vehicle is determined only by the edge information.

The head position detecting means 12b is based on the differential image in the differential image storage section 13 and the original image after noise removal.
Determine the beginning of the vehicle. First, the road surface removing means 12
The vehicle width,
Since the vehicle length has been determined to some extent, the differential image of the area where the vehicle exists is projected on the X-axis (SE1). Next, in order to obtain edge information, the differential image and the binarized image of the original image after noise removal are logically ANDed for each pixel, and the value is projected on the Y axis. Then, the differential projection values are sorted for each pixel,
The coordinates located at the bottom of the image (on the camera side) are selected as the tentative head position in the order of the sorted data amount, up to the predetermined order (SE2).

The selected Y-axis coordinate is determined to be the tentative head position. From the coordinates, the Y-coordinate is determined within a certain range in the Y-coordinate and the X-coordinate in the process up to this point on the processed image. The luminance value of the original image is projected on the Y-axis in the vehicle width range (SE3). Then, the brightness projection values are sorted, and the Y-axis coordinates of the darkest brightness on the camera side are searched from the temporary head position within the predetermined order in the order of the darkest brightness (SE4).

In the search in SE4, when there is a darker portion on the camera side than the temporary head position, the Y-axis coordinate is determined as the vehicle head position (SE5). On the other hand, in the search in SE4, when there is no darker portion on the camera side than the temporary head position, the temporary head position is determined as the vehicle head position (SE6). Thus, the head position is determined (SE7). FIG. 13 is a diagram illustrating an image of the determined vehicle head position. In this figure, for the vehicle part,
It has multiple gradations as the original image. It is to be noted that the error diffusion processing is illustrated for convenience.

In the search processing of SE4, there is a possibility that the shadow of a tree growing on the road shoulder or the like may have an effect. However, when the shadow of the vehicle appears, the shadow of the vehicle is darker in brightness. Since it has an effect of removing disturbance, accuracy can be improved as compared with a method of detecting the head using only edge information. If the shadow of the vehicle is not visible, the shadow of the tree is very unlikely to be present on the road surface, and it is necessary to enter a certain range from the head of the vehicle to make a misjudgment. It can be said that the possibility of misjudgment as the head of the vehicle is extremely low.

Next, the vehicle width at the head of the vehicle is determined by the vehicle width detecting means 12c. The vehicle width detecting means 12c calculates the vehicle width of the vehicle head from the vehicle head position detected by the head detecting means 12b. As the vehicle width, a value obtained by reading out the differential image from the differential image storage unit 13 and projecting the X-axis is used. This is because the horizontal differential value is difficult to represent the vehicle width. In order to reduce the influence of the shadow, when there is a shadow at the head position of the vehicle in the original image of the target coordinates, the brightness of the shadow is examined. If there is no brightness equal to or higher than a certain value based on the brightness of the shadow obtained at that time, a threshold value for setting the differential value to 0 is determined.

The logical product of the differentiated image and the binarized processed image is calculated in a certain range (for example, 3 m) upward (in the Y-axis direction) with respect to the vehicle width range obtained in the process up to this point from the vehicle head position. Take. The value is X-projected with an upper limit with the influence of the shadow removed. The upper limit at the time of projection is to remove the influence of the high brightness edge. With respect to the obtained projection value, the vehicle width is reduced until coordinates exceeding the threshold appear from the left and right of the image. Then, the coordinates exceeding the projection value threshold are determined as the vehicle width.

Note that there may be a case where the vehicle width edge is on the white line (masked by the filter unit 11) and the vehicle width cannot be obtained by this method. In this case, the vehicle width is determined using the Y-axis edge information. At this time, the vehicle width becomes shorter in length at the white line, so the accuracy may decrease.However, if the threshold of the vehicle width of large and small vehicles is separated by more than the width of the white line, the accuracy may be reduced. It is thought that the decrease can be reduced.

Next, the rear part of the vehicle is detected by the rear detecting means 12d. Since a large vehicle has a longer vehicle length than a small vehicle, a large or small vehicle type is temporarily determined using vehicle width information obtained so far. Based on the determination, the rear position of the vehicle is estimated.

In the case of a small car, there is a possibility that there is a following vehicle, so that processing is performed. The following vehicle may be present at the same rear position as the vehicle width position of the vehicle type discriminating vehicle in a plane (2D) image from the relation of the camera position, but it can be said that it is small, so each of the Y-axis lines The center of gravity of the differential value is used, and the center of gravity is compared with the center of gravity at the head of the vehicle.

Since the above processing alone is not sufficient for cutting out the following vehicle, the following vehicle is also judged from the distribution state of the differential value projected on the Y axis. After removing the following vehicle portion, the vehicle rear position is determined by using only the edge information method of the method of obtaining the vehicle head portion. In the case of a large vehicle, there is no following vehicle on the processing screen, so there is no need to consider the following vehicle.
FIG. 14 is a diagram showing an original image of a large vehicle. As shown, in the case of a large vehicle, the rear portion of the vehicle is off the upper part of the image.

Next, each part of the vehicle is determined by the vehicle part determining unit 15 (SP4 in FIG. 5). First, the silhouette extracting unit 15a performs an extracting process on the silhouette of the vehicle. FIG. 15 is a flowchart showing the process of silhouette cutout.

In the process up to this point, the leading position of the vehicle and the trailing position of the leading vehicle width have been known. However, since the rear part of the vehicle is farther from the camera at the rear and the edge is hard to detect, it is possible that the vehicle width does not indicate an accurate vehicle width. At that time, the vehicle width at the rear part of the vehicle is smaller than the original vehicle width. Further, from the conditions captured by the camera, the rear portion of the vehicle looks smaller and the front of the vehicle looks larger. When detecting the silhouette of the vehicle, only the portion where the silhouette from the rear to the front of the vehicle is large in terms of the vehicle width is effective in order to reduce the influence of the pattern of the vehicle body. An edge is detected for each line from the rear of the vehicle to the front of the vehicle within a range of the front vehicle width, and an edge having a predetermined value or more and its continuity are determined, and a silhouette is obtained based on the edge.

First, the differential image is read from the differential image storage unit 13, and the rightmost and leftmost edges of the effective edge from the rear part of the vehicle to the head of the vehicle are searched for each line in the Y-axis direction within the section of the vehicle width. Perform (SG). Specifically, a search is made for each line in the Y-axis direction from the rear part of the vehicle to the head of the vehicle within the section of the head vehicle width (SG1). Next, it is determined whether or not there is a line in which an edge (differential value) of a certain value or more is continuous with a length of a certain value or more (SG2). SG1 if NO
To find the next one line. If YES, the right and left ends of the continuous line are checked (SG3). Compare to the right and left edges found on the previous line, and if the right and left edges of the line are further to the right and left (that is, outside the vehicle), the right and left edges of the line A judgment is made (SG4). Thereafter, the process returns to SG2 to search for the next Y line.

Next, the leftmost and rightmost edges of the edge effective as a silhouette from the rear part of the vehicle to the head of the vehicle are examined. If the width of the Y-axis line increases in the direction of the head of the vehicle, the line is regarded as a silhouette. Judge as a valid edge (SH). Specifically, an effective edge from the rear part of the vehicle to the head of the vehicle is checked (SH1). It is determined whether or not the edge is a valid edge (SH2). If NO, the process goes to SH1 to check the next valid edge. In the case of YES, it is determined whether or not the vehicle width in the line increases toward the front of the vehicle (SH3). In the case of NO, shift to SH1 and thereafter,
In the case of YES, it is determined that the line is valid as a silhouette (SH4).

Next, from the rear part of the vehicle to the head of the vehicle, processing is performed to connect the right end and the left end of the effective edge of the silhouette (S
I). FIG. 16 is a diagram illustrating a silhouette cutout state. As shown in FIG. 4B, the head and rear positions of the vehicle are detected in the Y-axis direction. Also,
At the head position and the rear position of the vehicle, the positions of the left end and the right end of the vehicle are detected in the X-axis direction (width direction), respectively. Also, the rear position of the vehicle, the left end of the rear position and the left end of the head position, the right end of the rear position and the right end of the head position,
A silhouette in which the leading edge is connected is detected.

Next, each part of the vehicle is determined by the part determining means 15b. Since the silhouette does not always indicate the width of the vehicle at that position, the vehicle width information uses the information of the vehicle width at the leading position of the vehicle. Based on the coordinates detected as the edge where the vehicle width is widened at the time of silhouette cutout, a process of dividing the parts of the vehicle using the ratio of the vehicle length and the vehicle width is executed. Further, for a vehicle in which some edges are difficult to be detected, it may be conceivable that a part cannot be obtained from the information at the time of silhouette cutting. Here, the leading vehicle width of the vehicle is determined by the vehicle width detecting means 12c. At the same time, the vehicle length is also required based on the vehicle width.

Then, a differential image is read from the differential image storage unit 13 for the silhouette portion, and this differential image is
Axis projection. From the result, the part is determined using the ratio of the vehicle width and the vehicle length.

FIG. 17 is a diagram showing an image after the part of the vehicle has been determined. FIG. 18 illustrates the determination state of each part in the vehicle of FIG. As shown in these figures, the vehicle has a vehicle width in the X-axis direction and a predetermined height in the Y-axis direction (light portion,
(Corresponding to the height of the bonnet). At the upper position of the part of the vehicle head, a part (bonnet) from the vehicle head to the lower part of the windshield is determined. At the upper position of the bonnet part, a part (front glass) from the lower part of the windshield to the upper part of the windshield is determined. At the upper position of the windshield, a portion (roof) from the upper portion of the windshield to the rear of the vehicle is determined.
Each part is converted into mm based on the number of pixels on the X axis and the Y axis,
The actual length (width, height) and area are determined simultaneously.

Here, as shown in FIG. 18, it is possible to perform more accurate part determination by using a silhouette edge to determine a part than to determine a part based only on differential data. It has been shown.

The vehicle width and each part determined as described above are output to the vehicle type determining unit 17. The vehicle type determination unit 17
The vehicle type is determined based on these data and the feature data stored in the feature data memory 20 (SP5 in FIG. 5). Although the processing is performed in mm units internally, the final threshold value is determined with reference to a prepared sample.

FIG. 19 is a diagram showing characteristic data stored in the characteristic data memory 20. The parameters for determining the vehicle type are an apparent vehicle length, a vehicle width, a hood height, a hood position, a windshield position, and an area of each part.

First, it is determined whether the vehicle is large or small (SP51). If the vehicle width is larger than a certain value (for example, 3.0 m indicated by the feature data), it is determined that the vehicle is large (SP52). Similarly, the judgment of large or small size is made by using the detection area. This means that even if you make a mistake in detecting the vehicle width,
Execute to detect a large vehicle. If no large feature is detected, it is determined to be small (move to SP54).

Next, the vehicle determined to be large is further determined to be a cargo or a passenger (SP53). Compared to large passengers, large cargo has a smaller windshield, so their characteristics are compared and judged. Those with large windshields and a large ratio to the hood are considered large passengers, and the rest are judged as cargo.

On the other hand, if it is determined in SP51 that the small vehicle is small, it is next determined in SP54 whether the small vehicle is cargo or passenger. If the vehicle length is longer than the vehicle length ratio, it is determined to be a small cargo (SP55). Also, when the bonnet has a larger ratio than the vehicle length, it is determined to be a small cargo (SP56). Further, a bonnet having a high height is also determined to be a cargo (SP57). And
Those not determined as cargo (exclusion condition) are determined as small passengers (SP58).

The results of these vehicle type determinations are externally output to the display unit 5 and the like. The data is sequentially output to the number counting unit 30. The number counting unit 30 counts the number of detected vehicles for each vehicle type.

The vehicle type discriminating section 17 may be configured to accumulate and store characteristic data of the vehicle in the characteristic data memory 20 every time the vehicle type is determined by vehicle detection. According to this configuration, the feature data can be increased over time, and the accuracy of vehicle type determination can be improved over time.

In the above embodiment, the vehicle is imaged by the television camera. However, since the vehicle type is determined by a still image, a digital camera capable of photographing a still image when the vehicle is detected may be used. Further, the camera is configured to photograph two lanes, but it is also possible to photograph one lane or a plurality of lanes and determine the type of vehicle passing through each lane.

In the above embodiment, the head detecting means 12b determines the vehicle head position by image processing. However, the present invention is not limited to this. That is, the head position of the vehicle may be detected using the above-described detector such as an ultrasonic wave or an infrared ray. In this case, the vehicle type discrimination described in the above processing procedure can be executed only by the processing after detecting the head position of the vehicle by this detector. Further, by tracing the head position of the vehicle from a plurality of input images using the head position detection means of the present embodiment, the speed of the vehicle can be measured.

[0092]

According to the present invention, the vehicle type can be determined based on this image simply by taking an image of the traffic flow of a moving body approaching from a distance with a camera. This makes it possible to determine the vehicle type with low cost and high reliability by using only one general-purpose camera. In addition, since the configuration is such that predetermined image processing is performed on the image of the camera and each part of the vehicle is determined to determine the vehicle type, a large number of vehicle types can be accurately determined. In particular, because the road surface is removed by projecting the differential image, even if the vehicle brightness is close to the road surface brightness or if there is a shadow other than the vehicle on the lane, the road surface can be appropriately removed and only the vehicle can be extracted. , Less affected by the weather. Also,
The process of determining the head position of the vehicle is such that after obtaining a temporary vehicle width in the differential image, the darkest line within a certain range at the bottom of the image is determined as the vehicle head position, and the vehicle head position is accurately determined. can do.

Further, when the edge of the differential image extends from the rear position of the vehicle to the lower part of the image, the left and right ends of the vehicle are determined and a silhouette up to the front end position of the vehicle is obtained. Can be accurately determined. In the vehicle type determination, by combining the information of the silhouette and the information of the differential image, the accuracy of the vehicle type determination can be further improved without being affected by disturbance components.
In addition, by determining the vehicle type by referring to the feature data memory, it is possible to accurately determine the vehicle type by referring to a large amount of determination information. If the feature data is configured to be accumulated each time the device determines the vehicle type, the accuracy of the vehicle type determination can be gradually improved over time after the device is installed.

[Brief description of the drawings]

FIG. 1 is a diagram showing a camera installation state of a vehicle type identification device of the present invention.

FIG. 2 is a diagram illustrating an image range captured by a camera.

FIG. 3 is a block diagram showing the overall configuration of the apparatus of the present invention.

FIG. 4 is a block diagram showing the internal configuration of the device shown in FIG.

FIG. 5 is a flowchart showing a procedure of an entire process of vehicle type determination according to the present invention.

FIG. 6 is a diagram showing an original image captured by a camera.

FIG. 7 is a flowchart illustrating a processing procedure of road surface removal.

FIG. 8 is a diagram showing an image obtained by differentiating an original image.

FIG. 9 is a diagram showing an image at the time of road surface removal.

FIG. 10 is a diagram showing an image obtained by removing a road surface from an original image.

FIG. 11 is a diagram showing an image when noise is removed.

FIG. 12 is a flowchart illustrating a processing procedure of a vehicle head position determination.

FIG. 13 is a diagram showing an image at the time of determining a vehicle head position.

FIG. 14 is a diagram showing an original image of a large vehicle.

FIG. 15 is a flowchart showing a processing procedure of silhouette cutout.

FIG. 16 is a diagram showing an image at the time of silhouette clipping.

FIG. 17 is a diagram showing an image at the time of part determination.

FIG. 18 is a diagram illustrating an image of a part.

FIG. 19 is a diagram showing the contents of a feature data memory.

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 1 ... Camera, 2 ... Processing part, 3 ... Image memory, 5 ... Display part, 6 ... Operation part, 10 ... Image input part, 11 ... Filter part, 12 ... Vehicle position determination part, 12a ... Road surface removal means, 1
2b: Head detecting means, 12c: Vehicle width detecting means, 12d ...
Rear detection means, 13: Differential image storage section, 15: Vehicle site determination section, 15a: Silhouette cutout section, 15b: Site determination section, 17: Vehicle type determination section, 20: Feature data memory, 30: Number counting section.

Claims (30)

[Claims] 1. A vehicle type discriminating apparatus for discriminating a vehicle type of a road traveling vehicle during daytime, comprising: a camera (1) for capturing an image of the traveling vehicle; and a vehicle for removing a road surface based on an image captured by the camera. A processing unit (2) for detecting a head position and determining a vehicle type based on a vehicle width and a silhouette of the vehicle at the vehicle head position. 2. The daytime vehicle type discriminating apparatus according to claim 1, wherein the processing unit (2) includes a road surface removing unit (12a) for differentiating the image and removing a road surface from the image based on the differentiated image. . 3. The daytime vehicle type discrimination device according to claim 2, wherein the road surface removing means (12a) projects the differentiated image in the X-axis and Y-axis directions to determine and remove the road surface. 4. The road surface removing means (12a) projects the differentiated image in the X-axis and Y-axis directions to obtain an average value of projection values of each axis, and calculates the average value in the X-axis or Y-axis. 3. The daytime vehicle type discriminating apparatus according to claim 2, wherein a range having a lower projection value is determined as a road surface and removed. 5. The road surface removing means (12a) projects the differentiated image in the X-axis and Y-axis directions to obtain an average value of projection values of each axis, and calculates the average value in the X-axis or Y-axis. 5. The method according to claim 4, wherein luminance values in a range having a lower projection value are collected and sorted in the order of the luminance value having the largest abundance, and a pixel having a luminance value in a predetermined range centering on a certain order or up to a certain value is determined as a road surface. Vehicle type discrimination device in the daytime. 6. The processing section (2) differentiates the image and removes a road surface from the image based on the differentiated image to obtain a temporary vehicle width and a candidate vehicle position. The candidate images are projected on the Y-axis, and Y-axis lines having a projection value equal to or greater than a predetermined value are collected to form a temporary head candidate group,
The daytime vehicle type discriminating apparatus according to claim 1, further comprising a head detecting means (12b) for judging a Y-axis line at the lowermost position of the image among the Y-axis lines of the temporary head candidate group as a temporary head position of the vehicle. 7. The head detecting means (12b) projects the vicinity of the Y-axis line obtained as the tentative head candidate on the Y axis, sorts the brightness projection values, and sorts the brightness projection values in order of darkest brightness value. 7. The system according to claim 6, wherein when a Y-axis line located at a lower part of the image from the tentative head candidate is detected within a predetermined order, the detected Y-axis line is determined as a head position of the vehicle. Vehicle type discrimination device in the daytime. 8. The processing unit (2) differentiates an image of the vehicle head position based on the detected vehicle head position and projects the differentiated image on the X axis, and obtains a projection value equal to or greater than a predetermined value from the left and right and the vehicle width. The daytime vehicle type discriminating apparatus according to claim 1, further comprising a vehicle width detecting means (12c) for judging. 9. The processing unit (2) performs differential projection of the section on the Y-axis with a length within a certain range based on a vehicle length set corresponding to the detected vehicle width, and performs a certain value. 9. The daytime vehicle type discriminating apparatus according to claim 8, further comprising a rear detecting means (12d) for judging a farthest Y coordinate having obtained the projection value as a rear position of the vehicle. 10. The rear detecting means (12d) shifts the center of gravity of the Y-axis vehicle width partial line at the head position of the vehicle and the center of gravity of the differential value of each Y-axis line in the X-axis direction. Is within a predetermined range, the determined rear position is determined to be the rear position of the vehicle. On the other hand, if the predetermined range is exceeded, the vehicle is determined to be a succeeding vehicle and is not detected as a vehicle rear. The daytime vehicle type discrimination device according to claim 9, wherein the candidate is detected as a rear portion of the vehicle. 11. A vehicle type discriminating apparatus for discriminating a vehicle type of a road traveling vehicle during daytime, comprising: a camera (1) for capturing an image of the traveling vehicle; and a head of the vehicle in the image based on the image captured by the camera. A vehicle position determining unit (12) that determines a position, a vehicle width, and a rear position; a vehicle part determining unit (15) that determines an existing position of each part of the vehicle based on the determination information of the vehicle position determining unit; A vehicle type discriminating apparatus in the daytime, comprising: a vehicle type determining unit (17) for determining a vehicle type based on information on each part of the vehicle determined by the vehicle part determining unit. 12. The vehicle position determining unit (12) differentiates the image based on an image captured by the camera,
Road surface removing means (12a) for removing a road surface from an image based on the differentiated image; a temporary vehicle width and a vehicle existing position of the vehicle by differentiating the image and removing the road surface from the image based on the differentiated image; A candidate is obtained, and an image of the existence position candidate is projected on the Y axis,
A Y-axis line having a projection value equal to or more than a predetermined value is collected and set as a tentative head candidate, and the Y-axis line at the lowermost position of the image among the Y-axis lines in the tentative head candidate group is determined as a tentative head position of the vehicle. Head detecting means (12b); means for judging a head position based on the temporary head position and a luminance projection value in the vicinity thereof; and a vehicle which differentiates an image of the vehicle head position based on the detected vehicle head position. A vehicle width detecting means (12c) for obtaining a width; and a rear detecting means (12d) for determining a rear position of the vehicle based on a vehicle length set corresponding to the detected vehicle width.
The vehicle type in the daytime according to claim 11, further comprising: a vehicle part determination unit (15) that determines each part of the vehicle based on information on a head position, a vehicle width, and a rear part position of the vehicle after the road surface is removed. Discriminator. 13. The vehicle part judging section (15) is connected to the left end of the left edge of the rear position of the vehicle and the right end of the right edge of the rear position detected based on the differential image obtained by differentiating the image in respective Y-axis coordinates. 12. The daytime vehicle type discriminating apparatus according to claim 11, further comprising a silhouette cutout means (15a) for judging an edge to be performed as a left end and a right end of the vehicle and cutting out a silhouette of the vehicle until reaching a head position of the vehicle. 14. The silhouette cutting means (15)
a) determines a line that is continuous with a certain or more differential value in the X-axis direction as a valid edge based on a differential value obtained by projecting the differential image in the Y-axis direction, and determines the leftmost edge among the valid edges at the rear of the vehicle. 14. The vehicle according to claim 13, wherein the effective edge, which extends at least toward the lower end of the image with respect to the left end and the right end of the rightmost edge, is determined as the left end and the right end of the vehicle, and the silhouette of the vehicle is cut out until reaching the leading position of the vehicle. Daytime vehicle type identification device. 15. The vehicle part judging section (15) sets a vehicle to a head portion (hood height), a hood portion, based on a ratio of a vehicle width to a vehicle length of the cut-out silhouette.
The daytime vehicle type discriminating apparatus according to claim 11, further comprising a part determining means (15b) for determining each part of a windshield portion and a rear portion after the roof. 16. The part determination means (15b), in addition to the part determination based on the silhouette, a vehicle width based on an edge obtained by projecting, in the Y-axis direction, a differential image obtained by differentiating an image captured by the camera. 16. The daytime vehicle type discrimination device according to claim 15, wherein each part of the vehicle is determined based on a ratio of the vehicle length and the vehicle length. 17. A feature data memory (20) in which feature data of a vehicle type is stored in advance, wherein the vehicle type determination unit (17) stores determination information of the vehicle part determination unit in feature data of the feature data memory. The vehicle type discrimination device in the daytime according to claim 11, wherein the vehicle type is determined with reference to the following. 18. The daytime vehicle type discriminating apparatus according to claim 17, wherein the vehicle type judgment unit (17) accumulates and stores the judgment information at the time of the vehicle type judgment in the feature data memory (20). . 19. The vehicle type determining unit (17) determines whether the vehicle is a large vehicle based on the vehicle width and / or the area of each part at the head position of the vehicle determined by the vehicle part determining unit (15). The daytime vehicle type discriminating apparatus according to claim 11, which determines whether the vehicle is a small vehicle. 20. The vehicle type judging section (17) judges whether the vehicle is a large cargo vehicle or a large passenger vehicle based on an area ratio between a windshield and a bonnet when the result of the determination is a large vehicle. 19. A daytime vehicle type discrimination device according to claim 19. 21. The vehicle type determining unit (17) determines whether the vehicle is a small cargo vehicle or a small passenger vehicle based on the ratio of the vehicle width to the vehicle length and / or the height of the hood when the determination result is a small vehicle. 20. The daytime vehicle type identification device according to claim 19, which determines whether there is a vehicle. 22. A vehicle type discriminating method for discriminating a vehicle type of a road traveling vehicle during daytime, comprising: capturing an image of a traveling vehicle with a camera; differentiating an image captured by the camera;
After projecting in the axis and Y axis directions to determine and remove the road surface,
A vehicle type discrimination method in the daytime, wherein a vehicle type is determined based on a vehicle width at a vehicle head position and a vehicle silhouette. 23. Projecting the differentiated image in the X-axis and Y-axis directions to obtain an average value of projection values of each axis, and determining a range having a differential value lower than the average value in the X-axis or Y-axis on a road surface. 23. The method of discriminating a vehicle type in the daytime according to claim 22, wherein: 24. Projection of the differentiated image in the X-axis and Y-axis directions to obtain an average value of projection values of each axis, and a luminance in a range having a differential value lower than the average value in the X-axis or Y-axis. Gather the values, sort them by most abundance,
23. The daytime vehicle type discrimination method according to claim 22, wherein a pixel having a luminance value within a predetermined range around each of the higher luminance values is determined as a road surface. 25. A tentative vehicle width of a vehicle and a candidate position of a vehicle are obtained by removing a road surface from the image based on the differentiated image, and the image of the candidate position is projected on the Y-axis, and 23. The Y-axis line having the projection value of (i) is collected and used as a temporary head candidate, and the Y-axis line at the lowermost position of the image among the Y-axis lines of the temporary head candidate is determined as the head position of the vehicle. Daytime vehicle type identification method. 26. Based on the Y-axis line obtained as the tentative head candidate, luminance projection is performed on the Y-axis within a certain range, and the luminance projection values are sorted. 26. The daytime vehicle type discriminating method according to claim 25, wherein when the Y-axis line located at the lower part of the image from the tentative head candidate is detected, the detected Y-axis line is determined as the head position of the vehicle. . 27. The vehicle width at the vehicle head position is obtained by differentiating the image and projecting the X axis in a certain range based on the vehicle head position and obtaining a projection value equal to or more than a certain value from the left and right.
2. The vehicle type discrimination method in the daytime described in 2. 28. A rear part position of the vehicle is obtained based on a vehicle length set corresponding to the detected vehicle width, and a center of gravity of a vehicle width partial line on the Y axis at a head position of the vehicle is obtained. When the deviation of the center of gravity of the differential value of each line in the X-axis direction is within a predetermined range, the determined rear position is determined to be the rear position of the vehicle, while when the deviation exceeds the predetermined range, 23. The daytime vehicle type discriminating method according to claim 22, wherein a next candidate satisfying the condition is detected as a vehicle rear part without detecting that the vehicle is a following vehicle as a vehicle rear part. 29. Based on a differential value obtained by projecting the differential image in the Y-axis direction, a line continuous with a certain differential value or more in the X-axis direction is determined as an effective edge, and a left edge of the effective edge at the rear of the vehicle is determined. The effective edge, which extends at least to the lower end of the image with respect to the right end of the left end and the right end of the right end, is determined to be the left end and the right end of the vehicle, and the silhouette of the vehicle is cut out until reaching the leading position of the vehicle. 23. The vehicle type discrimination according to claim 22, wherein each part of the vehicle is discriminated based on a ratio of a vehicle width and a vehicle length of the silhouette to a head portion (hood height), a hood portion, a windshield portion, and a rear portion after the roof. Method. 30. The vehicle is a large vehicle, a small vehicle, a large cargo vehicle, or a large passenger vehicle based on the determined vehicle width at the leading position of the vehicle and data of each part. 30. The daytime vehicle type determination method according to claim 29, wherein the method determines whether the vehicle is a small cargo vehicle or a small passenger vehicle.