JPS63174199A - Vehicle recognition system - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a vehicle recognition system for recognizing a vehicle running or stopped on a road surface.

[Conventional technology]

Various types of traffic flow measurement systems using image processing have been known. Specifically, for example, JP-A-6
Publication No. 1-214100, Traffic Wi using image processing method
``Research on Traffic Flow Measurement Technology Using Image Processing Methods'' (March 1982) published by T-Measurement Technology Research Group and r published by the same research group.
"Traffic flow measurement system using CCD image sensor"
(March 1982), etc.

The conventional method will be explained below.

For example, CC is used to photograph the road surface and vehicles running on the road surface.
It is done by a TV camera. The generation of image patterns by the camera was performed by dividing the patterns into lanes, assuming that the vehicle was traveling in accordance with the lane divisions. The image signal thus obtained is digitized by the slave CPU, and the vehicle is recognized by the next master CPU. That is, by comparing the brightness of the road surface and the brightness of the vehicle (vehicle body), the vehicle was recognized from the difference in brightness.

[Problem that the invention seeks to solve]

However, the above conventional method has the following problems.

First, in the conventional method, vehicle recognition was performed by detecting the vehicle's brightness and comparing it with the road surface brightness (reference brightness). were significantly different, making accurate recognition difficult.

Second, in the conventional system, vehicle recognition was performed only on a portion of the lanes on the road. Therefore, if a vehicle rarely travels across different lanes, or if a vehicle rarely changes lanes, it is difficult to recognize the vehicle, and the same vehicle may be recognized more than once. Roughly,
When different vehicles are traveling in two or more lanes, it is difficult to recognize them separately.

Therefore, a first object of the present invention is to provide a vehicle recognition method that can accurately recognize a vehicle regardless of the vehicle type (1'), paint color, etc.

Moreover, the second thing of the present invention straddles two lanes,
A vehicle recognition method that can accurately recognize vehicles that are changing lanes or when two or more vehicles are running parallel to each other in multiple lanes. The purpose is to provide

[Means for solving problems]

The first vehicle recognition method according to the present invention detects pattern portions with different brightness from an image pattern obtained by shading the road surface and the vehicle.
The first method detects a pattern portion of a road surface portion whose luminance is lower than the base luminance as a shadow portion of the vehicle.
a second step of detecting a pattern portion of a predetermined brightness formed on the opposite side of the road surface portion across the shadow portion as a vehicle portion; and a shadow portion of the vehicle detected in the first and second steps. and a third step of recognizing the presence of a vehicle based on the vehicle part.

The second vehicle recognition method according to the present invention identifies pattern portions with different brightness from an image pattern obtained by shadowing the entire lane area of the road surface and the vehicles traveling on the entire lane area of the road surface. This is a method for recognizing vehicles by detecting them, and the first step is to detect a pattern part of the road surface whose brightness is lower than the standard brightness as a shadow part of the vehicle, and a plurality of shadow parts exist on the road surface in the lane and cross direction. In this case, the second step is to identify a plurality of shadow parts for each vehicle, and the second step is to identify a pattern part of a predetermined brightness that appears on the opposite side of the road surface part across the shadow parts identified for each vehicle in the second step. a third step of detecting the vehicle part as a vehicle part identified for each vehicle; a shadow part of each vehicle detected in the first step and identified in the second step; and a vehicle part detected in the third step. and a fourth step of recognizing the vehicle based on each vehicle part.

[Effect]

The first vehicle recognition method according to the present invention is configured as described above, so that a vehicle on a road surface can be detected by detecting a shadow part of a vehicle with a luminance lower than the standard R degree of the road surface, and detecting a part of the vehicle with a predetermined luminance. It acts to recognize its existence by detecting it.

The second vehicle recognition method according to the present invention is configured as described above, so that vehicles running on the road surface can be identified one by one by detecting a shadow part with a luminance lower than the standard luminance of the road surface. Therefore, by detecting the vehicle part having a predetermined brightness following the shadow part, it is possible to recognize each vehicle.

〔Example〕

Hereinafter, some embodiments of the present invention will be described with reference to the accompanying drawings. In the description of the drawings, the same elements are denoted by the same reference numerals, and redundant description will be omitted. FIG. 1 is a flowchart of an embodiment of the method of the present invention, FIG. 2 is a block diagram showing the basic configuration of a vehicle recognition device to which the embodiment is applied, and FIG. 3 is a block diagram of the device shown in FIG. FIG.

As shown in FIG. 2, the vehicle recognition device includes a VTR 2 connected to a fixed camera 1, an image processing device 3 that processes image signals from the VTR 2, a monitor television 4 that monitors images from the VTR 2, and an image processing device. 3, and an operation console 5 for operating the 3.

The image processing device 3 also includes an image memory 7 for digitizing the analog image signal from the VTR 2, a slave CPU 8 for digitizing the digital image signal, and a ternarizing signal from the slave CPU 8 to control the vehicle. It has a master CPU 9 for recognition, etc. In general, the VTR 2 not only receives an image signal from the fixed camera 1 via a signal line, but also receives an image signal from a portable camera 6 in the form of a magnetic tape or the like.

Next, the detailed configuration of the image processing device 3 will be explained based on FIG. As shown in the figure, the image memory 7 has an A/D converter 11 for A/D converting an analog image signal from the VTR 2. The A/D converted digital image signal is provided to a ternarization calculation unit 14 via a pixel extraction unit 12 and a luminance difference calculation unit 13. Then, the ternarized signal is transmitted to the front shadow detection section 15 and the bonnet detection section 16.
and the roof detection section 17, and these detection results are sent to the next master CPU 9. The detection results given to the master CPU 9 are first input to the vehicle recognition section 21, where the vehicle is recognized.

Further, the results are sequentially sent to the vehicle movement tracking unit 22, lane determination unit 23, vehicle measurement unit 24, and measured value output unit 25, and the results are sent to the console 5 and the recognition results are displayed on a CRT screen or the like. It will be done.

Next, with reference to FIG. 4, the functions and actions of each component of the image processing device will be explained. ) to digital conversion. In this digitization, if A/D conversion is performed, for example, at 8 bits/pixel, the resulting digital signal will be luminance data with 28=256 gradations. This image data is then given to the pixel detection section 11 of the slave CPU 8 via a signal line.

The pixel detection unit 12 extracts luminance data for each pixel corresponding to a measurement point in the measurement area. Here, the measurement area is set, for example, as shown by reference numeral 31 in block P1 in FIG. 4, and measurement points 32 are arranged in a grid pattern in this measurement area 31. Therefore, the luminance data at each measurement point 32 in the measurement area 31 is sent from the pixel extraction section 12 to the luminance calculation section 13. An example of this luminance data is shown in block P2 of FIG.

The brightness information calculation unit 13 calculates the difference (difference value) between the brightness data of each pixel (measurement point 32) and the reference brightness of the road surface. That is, the road surface usually has a substantially constant sound intensity, and this brightness (road surface reference brightness) does not change at certain time intervals and is fixed depending on the measurement area. Therefore, this reference brightness is determined in advance or at any time, and compared with the brightness data of each measurement point. Then, for example, a pixel whose luminance is higher than a certain level than the road surface...+1 a pixel whose luminance is approximately the same as that of the road surface...O a pixel whose luminance is lower than the road surface by a certain level or more...-1 is converted into three values. The ternary data is then provided to the front shadow detection section 15, the hood detection section 16, and the roof detection section 17. Note that this ternarized data is shown in block P3 in FIG.

The foreshadow detection unit 15 detects a pattern portion darker than the road surface (a portion with ternary data = -1). Then, the continuity of this dark pattern in the vertical direction (the direction in which the vehicle is traveling) and the horizontal direction (the direction across the path) is checked, and if it is continuous over a certain range, it is determined that it is the front shadow of the vehicle. For example, in block P3 of FIG. 4, a group of ternary data = -1 appears in the left side of the measurement area, and it can therefore be seen that the foreshadow of the next vehicle is present in this area.

As a result, the position of the front shadow of the vehicle is detected, and the result is sent to the vehicle recognition unit 21 of the master CPU 9.

The hood detection unit 16 detects the hood of the vehicle that appears following the front shadow. Here, when the vehicle is running toward the camera, the bonnet should appear next to the front shadow of the vehicle, and its brightness is different from the front shadow, being higher in brightness than the road surface. There are many cases. Therefore, what is necessary is to check the group of 31a data that is continuous with the direction of the foreshadow. For example, in block P3 in FIG. 4, a bonnet portion with ternary data = +1 appears following the foreshadow on the right side of the measurement area.

The roof detection section 17 detects the roof section of the vehicle that appears following the hood section. For example, in block P3 in FIG. 4, a roof appears in the center of the measurement area. The detection results of the bonnet detection section 16 and the roof detection section 17 are as follows.
Similar to the detection result of the foreshadow detection section 15, the detection result is sent to the vehicle recognition section 21.

The vehicle recognition unit 21 recognizes that the vehicle is a single vehicle based on the detection results of the front shadow, bonnet, and roof. That is, it is checked whether the patterns of the front shadow, bonnet 1~, and roof are within a certain range in the vehicle width direction, and if they are within a certain range, one vehicle is determined.

The vehicle movement tracking unit 22 assumes that a vehicle detected at regular time intervals is stopped or moving forward, and tracks the vehicle. That is, vehicles that are being tracked are not included in the number of vehicles, and only vehicles that have newly started tracking are counted.

The lane determining unit 23 determines in which lane of the road the center (center in the vehicle width direction) of each detected vehicle is located. The vehicle measurement unit 24 counts vehicles classified by lane for each lane. The count output unit 25 outputs data such as the number of vehicles traveling in each lane to the console 5.

Next, the operation of the system of the present invention will be explained with reference to FIGS. 1 and 3 to 5.

The analog image signal from the VTR 2 is sent to the A/D converter 11.
For example, it is converted into digital luminance data of 256 gradations (step 102). Then, the pixel extraction unit 12
After the luminance data for each pixel is extracted, the luminance calculation unit 13 calculates the difference (difference value) from the road surface reference luminance for each pixel (step 103). The difference value data for each pixel obtained in this way is sent to the ternarization calculation unit 14.
r+1J (bright), "O" (medium),
It is converted into r-1J (dark) ternarized data (step 106).

Next, 311! The i-ized data is sent to the foreshadow detection section 15,
Here, it is detected whether the ternary data of each pixel is darker than the road surface (step 108). That is, in the case of the embodiment, since the ternarized data of the road surface portion is "O" as shown in block P3 of FIG. 4, the ternarized data is "-".
1" will be detected. Then, it is checked whether the dark areas detected in this way are continuous in the vertical and horizontal directions (step 110). At this time, for example, if the measurement area 31 is set as shown in FIG. 5(A), a plurality of vehicles may be lined up laterally as shown in FIG. 5(B). However, the front shadows of these vehicles are rectangular and are spaced apart from each other, as shown in black and marked S in FIG. 5(B). Therefore, a rectangular group of dark parts can be recognized as the foreshadow of the vehicle (step 112).

The ternary data is also sent to the bonnet detection section 16 and the roof detection section 17. Then, the bonnet and the roof are detected (step 114).

Here, assuming that the vehicle is running on the road toward the camera or is stopped, the hood will always appear behind the front shadow, and the roof will always appear behind the hood. . Further, the brightness of the hood and roof is higher than the brightness of the front shadow, and is often higher than the brightness of the road surface. However, the brightness is not constant compared to the brightness of the road surface. Therefore, the bonnet and roof of the vehicle can be detected by checking whether there is a pattern with the highest brightness next to the foreshadow or a pattern with disordered brightness.

The detection results by these detection units 15 to 17 are sent to the next vehicle recognition unit 21. And the front shadow of the detected vehicle,
A determination is made as to whether the bonnet portion and the roof portion appear within a certain range in the lateral direction (step 116). In other words, as shown in Figure 5(B), the front shadow, bonnet, and roof of each vehicle should appear within a certain range in the lateral direction, so that it is possible to recognize them as one vehicle. You will be able to do it (step).

After the above processing is completed, vehicle tracking, lane Q determination, vehicle measurement, etc. are performed. The measurement results are output and displayed on the CRTk: of the console 5.

The present invention is not limited to the above embodiments, and various modifications are possible.

In the embodiment, the shadow under the bumper of the vehicle is detected, but the invention is not limited to this.

For example, a moving vehicle may be reflected by a camera from behind, and a shadow formed at the rear of the vehicle may be detected, or a shadow formed at the side may be detected. In general, the vehicle is not limited to one that runs on a road surface, but may be one that is parked on the roadside or in a parking lot.

As the camera, for example, a CCTV camera can be used, but this is not the only option, and it is not essential to use a VTR. In other words, it is also possible to directly connect the camera and the image processing device and perform online real-time processing.

The brightness data is not limited to 28=256 gradations, but 24
= 16 gradations, 26 = 64 gradations, etc. may be used. Furthermore, the brightness data is not limited to ternary data, and may be any type of data such as 4-value or 5-value data.

In the embodiment, the hood and roof of the vehicle are detected, but the detection is not limited to this. For example, the bonnet and the roof may be detected all at once, a part of the bumper may be detected in addition to the bonnet and the roof, and in some cases detection of headlights, etc. may be combined. .

〔Effect of the invention〕

As explained in detail above, according to the first vehicle recognition method according to the present invention, the shadow part of the vehicle that appears with almost constant brightness is detected regardless of the vehicle model, paint color, etc. Since this is used to recognize the vehicle, it is possible to accurately recognize the vehicle.

Further, according to the second vehicle recognition method according to the present invention, the shadow portion of the vehicle is detected across a plurality of lanes, and this is used in combination with the detection of the hood etc. while identifying each vehicle. Now, even if there is a vehicle traveling across two lanes or changing lanes, or when two or more vehicles are running parallel to each other in multiple lanes, these vehicles will be recognized. It has the effect of accurately recognizing the

[Brief explanation of the drawing]

FIG. 1 is a flowchart illustrating an embodiment of the method of the present invention, FIG. 2 is a block diagram showing the basic configuration of a device to which the embodiment method of FIG. 1 is applied, and FIG. 3 is a summary of the device of FIG. FIGS. 4 and 5 are block diagrams showing detailed configurations of the parts, and are explanatory diagrams of the embodiment system shown in FIGS. 1 to 3. 31...Measurement area, 32...Measurement point. Basic block 2 of the device to which the method of the present invention is applied
figure

Claims (1)

[Claims] 1. A vehicle recognition method that recognizes a vehicle by detecting pattern portions with different brightness from an image pattern obtained by photographing a road surface and a vehicle existing on the road surface, a first step of detecting the pattern portion with a brightness lower than a reference brightness of the road surface portion as a shadow portion of the vehicle; and the pattern having a predetermined brightness formed on the opposite side of the road surface portion with the shadow portion in between. a second step of detecting a part as a vehicle part; and a third step of recognizing the presence of the vehicle based on the vehicle's shadow part and the vehicle part detected in the first and second steps. A vehicle recognition method characterized by: 2. The vehicle recognition system according to claim 1, wherein the vehicle travels in a fixed direction on the road surface. 3. The vehicle recognition system according to claim 1, wherein the shadow portion is a front shadow portion formed below a bumper of the vehicle. 4. The vehicle recognition system according to claim 1, wherein the vehicle is photographed by a television camera, and the image pattern is a digital image pattern obtained by digitizing the image photographed by the television camera. 5. The vehicle recognition system according to claim 4, wherein the digital image pattern is obtained by ternarizing the photographed image. 6. The vehicle recognition method according to claim 1, wherein the second step includes the steps of detecting a bonnet portion of the vehicle and detecting a roof portion of the vehicle. 7. In a vehicle recognition method that recognizes a vehicle by detecting pattern portions with different brightness from an image pattern obtained by photographing the entire lane area of the road surface and a vehicle traveling on the entire lane area of the road surface. , a first step of detecting the pattern portion having a luminance lower than a reference luminance of the road surface portion as a shadow portion of the vehicle; when a plurality of the shadow portions exist on the road surface in a direction transverse to a lane, the plurality of shadow portions are detected; a second step of identifying a shadow portion of each vehicle in the second step; and a pattern portion of a predetermined brightness formed on the opposite side of the road surface portion across the shadow portion identified for each vehicle in the second step. a third step of detecting the vehicle part as the vehicle part identified for each vehicle; a shadow part of each vehicle detected in the first step and identified in the second step; and a fourth step of recognizing the vehicle based on the vehicle part of each vehicle detected in the step. 8. The vehicle recognition system according to claim 7, wherein the shadow portion is a front shadow portion formed below a bumper of the vehicle. 9. The vehicle recognition system according to claim 7, wherein the vehicle is photographed by a television camera, and the image pattern is a digital image pattern obtained by digitizing the image photographed by the television camera. 10. The digital image pattern divides the photographed image into 3
The vehicle recognition method according to claim 9, which is obtained by converting the vehicle into a value. 11. The vehicle recognition method according to claim 7, wherein the third step includes the steps of detecting a hood portion of the vehicle and detecting a roof portion of the vehicle.