JPH0238979B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a vehicle number image matching device used to measure travel time of a vehicle.

The traffic control systems that are maintained mainly in major cities across the country have various control measures using software, but the method of measuring the vehicle travel time, which is the evaluation index, is by using imaging cameras such as TV cameras. A method has been proposed in which the vehicle number image of a vehicle is photographed with a camera and the vehicle number images obtained by the same vehicle passing through two points A and B are compared with each other. This technology is an image processing technology that combines an imaging camera such as a TV camera with a microprocessor, and as a substitute for the human visual system, it enables object recognition that was impossible with conventional sensors. However, if an attempt is made to realize a specific device based on this technology, the scale will increase in terms of both hardware and software, making it difficult to realize.

SUMMARY OF THE INVENTION Therefore, an object of the present invention is to provide a highly practical vehicle number image matching device that can reliably match vehicle number images with a simple device configuration.

In order to achieve the above object, the vehicle number image matching device according to the present invention is a device for matching the vehicle numbers of the same vehicle passing through at least two points, based on an image signal obtained by an imaging camera. In addition to determining the center line, a two-dimensional partial image including a vehicle number image defined by a vehicle height line located at a certain distance from the front end of the vehicle and vehicle width lines to the left and right of the center line is cut out. A part that converts a two-dimensional partial image into a one-dimensional serial signal, and a part that converts the two-dimensional partial image into a one-dimensional serial signal, and standardizes the field machine geometry caused by the imaging of the imaging camera to this one-dimensional serial signal, and eliminates the deviation of the center line and the deviation of the vehicle height line. so that they are aligned with each other, and 2
It is characterized by having a part that performs cross-correlation on the serial signals of the points.

DESCRIPTION OF THE PREFERRED EMBODIMENTS The content of the present invention will be specifically described below with reference to the accompanying drawings, which are examples. FIG. 1 is a block diagram of a car number image matching device according to the present invention. In the figure, reference numerals 1 and 6 denote imaging cameras, which in this embodiment are comprised of TV cameras with shutters. These imaging cameras 1 and 6 are placed at point A and point B, respectively, which are separated from each other by an appropriate distance.

At the subsequent stage of the imaging cameras 1 and 6, there is a video amplifier 2 or 7 that amplifies the video signal provided from the imaging camera, and a video amplifier 2 or 7 that converts the analog video signal provided from the video amplifier 2 or 7 into a digital video signal and stores it. A/D conversion, image memory 3
or 8, the processing device 4 or 9 is connected, respectively. 5 and 10 are transmission devices for transmitting vehicle number image information at point A to point B through line 13;
1 is a correlator that stores the car number image information at point B given by the processing device 9 and the car number image information at point A given from the transmission device 10, and calculates the cross-correlation of both information; 12 is the correlator in the correlator 11; This is a control circuit that controls the timing of cross-correlation.

Next, each function of each of the above-mentioned constituent parts will be explained.

First, image information processing of the vehicle by the imaging cameras 1 and 6 will be explained. FIG. 2 shows the image information. L is a detection line set in the vehicle width direction on the road surface, and Si is a sample point. Imaging camera 1
If 6 and 6 are constituted by ordinary scanning TV cameras, such a still image will be obtained at a rate of one every 1/30 seconds in the case of a low-speed vehicle. For high-speed vehicles, a shutter with limited exposure time is provided.
Obtained by TV camera.

FIG. 3 shows the arrangement of the imaging cameras 1 and 6 and a three-dimensional view of the monitoring field for one-lane monitoring,
13 is an arm that supports the imaging camera 1 or 6. Figure 4 is a side view of the stereoscopic view of the monitoring field in Figure 3, and Figure 5 is the same view from the plane, α
is the horizontal angle of view, β is the vertical angle of view, γ is the vertical deviation angle, C is the distance to the actual size monitor screen, and h is the installation height of the imaging camera. Figure 6 shows the monitor screen, and the screen size is 2Ctan (β/2) x 2Ctan
(α/2). The pixel resolution of the monitor screen is
If it is 512×512, the ground height of arm 14 is 5
m, and if the lane width is 3 m, it is 0.6 cm/1 pixel.

In FIG. 3, the following relational expression holds between the ground surface coordinates (Xg, Yg) and the monitor screen coordinates (Xm, Ym).

Xm=(Xg・C・sinγ)/(Yg+htanγ) Ym=Ccosγ−hCsinγ/(Yg+htanγ) Here, when the vehicle enters the detection line L,
When Yg/h is sufficiently small, the amount of movement ΔYm in the Y direction becomes ΔYm=Ccosγ·(Yg/h). As shown in Figure 6, the monitor screen size is 2Ctan (β/2) x 2Ctan (α/2), so the relative movement amount on the monitor screen is △Ym/{2Ctan (β/2) )} = {cosγ・Yg/h}/{2tan(β/2)}. If a high-speed vehicle traveling at 60 km/h is 1/500
When photographing with an imaging camera with an exposure time of seconds, the vehicle moves by 3 cm/(1/500) seconds, and the corresponding number of pixels on the monitor screen is 5 × 10 ^-2 × 512 =
There are 2 pixels. However, in Figures 2 to 5, h = 5
m, L1=15m, and L2=5m. In this way, it is possible to obtain a car number image of a high-speed vehicle that is generally close to the original image.

Next, we will discuss the functions and algorithms of each part for matching the still images of the vehicle number images of the same vehicle obtained at two points A and B as described above in Section 7.
This will be explained with reference to the car number verification concept flow shown in the figure.

(B) Vehicle detection and image storage In Fig. 2, multiple sample points Si are installed on the detection line L of the monitor screen, and the brightness values of these sample points Si are normally transmitted through the video amplifier 2 or 7 to A/ By inputting it to the D conversion/image memory 3 or 8, A/D conversion (analog-digital conversion, the same applies hereinafter) is performed, and by taking the difference between the A/D conversion signal of the road brightness value of other parts, Detect vehicles. At this time, also in the imaging camera 1 or 6, one photographic image is obtained in one frame (usually 1/30 sec) as a standard NTSC signal.
In this way, once a vehicle is detected on the detection line L, in the next frame, the image in the shaded area in Figure 2 is A/D converted at high speed in the A/D conversion/image memory 3 or 8. Store brightness value. If the number of pixels in this shaded area is 512 x 100, then A/D
If a 3-bit, 10MHz conversion speed is used as the video A/D converter constituting the conversion/image memory 3 or 8, the conversion time will be 512 x 100 x 0.1 μsec.
= conversion speed of 5.12 msec, which is sufficient for practical use. Furthermore, the image memory capacity is about 20K bytes, which is sufficient and can be handled satisfactorily with current technology.

In the vehicle detection and image storage described above, the image signal in the shaded area obtained at the time of vehicle detection is a value obtained by quantizing the luminance value. The car number image originally does not have continuous density, but is a binary original image, but considering dirt, distortion, changes in environmental illuminance, etc. of the car number image, it is not converted into a binary image and is converted into a 3-bit image. It is more convenient to use quantized values. Note that the apertures of the imaging cameras 1 and 6 are automatic apertures, and the video amplifiers 3 and 8 are equipped with an automatic gain control function.

(b) Vehicle leading edge determination, vehicle center line determination, and partial image extraction.

Each of these processes is performed by a processing device 4 or 9 connected after the A/D conversion/image memory 3 or 8. First, the image signal of the shaded area quantized as a brightness value is input to the processing device 4 or 9, and the difference from the road brightness value is calculated.
As shown in FIG. 8, a partial image of only the vehicle is obtained. By performing spatial smoothing processing such as noise removal on this image signal, the coordinates x1, x2 of both ends of the vehicle and the coordinate y1 of the front end of the vehicle can be determined. The center line xo of the vehicle can be calculated from xo=(x1+x2)/2.

Next, the vehicle center calculated as described above
A partial image including the car number is cut out based on xo and the vehicle tip coordinate y1. The partial image including the car number is the 8th
In the figure, it is a two-dimensional partial image surrounded by (xo-ΔH), (xo+ΔH), y1, and (y1+ΔV). This partial image can be easily obtained by specifying an address in the image memory. The part that includes the car number, that is, the number plate attached to the front of the vehicle, is located at a height of 30 to 40 cm above the ground for regular cars and 60 to 70 cm for large cars, and is approximately constant. . Also, the deviation from the vehicle center xo is within ±1 cm. What's more, the size of license plates is not infinite; most are around 10cm x 20cm. Therefore, ΔH for cutting out the partial image
The values of and ΔV can be easily determined.

(c) One-dimensional conversion of two-dimensional partial images Figure 9 is a diagram showing the two-dimensional image memory of the above partial images, where the luminance values of the partial images are stored at addresses (xo, yo) to (xn, yn). There is. Therefore, this two-dimensional image memory is (xo, yo) → (xn, yo) →
(xn, yo) → (xo, y1) → (xn, y1) →……→
If you read them out sequentially and arrange them like (xn, yn),
It can be converted into a one-dimensional signal as shown in FIG. In this way, the one-dimensional signal of the partial image obtained at point A is transmitted to point B through the transmission device 5 and line 13, and is input to the cross-correlator 11 through the transmission device 10 provided at point B.

(d) Standardization and cross-correlation of vehicle number photography The cross-correlator 11 installed at point B allows
By cross-correlating the one-dimensional signal of the partial image obtained at point B and the one-dimensional signal obtained at point B, it is determined whether the vehicle number images of the same vehicle match or do not match. is: a Standardization of vehicle number photographing size at points A and B b Calculation error of vehicle center xo at points A and B c Calculation error of vehicle tip y1 at points A and B Within about 1 second,
Real-time cross-correlation decisions must be made. Here, it is assumed that the vehicle travels in a direction perpendicular to the detection line L, the horizontal angle of view is equal on the left and right sides, and the photographed image is taken from the front of the vehicle. 1st
FIG. 1 shows a block diagram of the cross-correlator 11.
In FIG. 11, 111 and 112 are storage units that store one-dimensional signals at point A and point B, respectively;
A D/A converter 115 converts the digital signal from 2 into an analog signal, and 115 is the D/A converter 113.
and an analog multiplier that performs multiplication of the analog signal outputs of 114; 116 is the analog multiplier 115;
117 is a processing circuit that determines the correlation based on the comparison output given from the comparator 116; 119;
is a storage device that stores a two-dimensional partial image of point B. Next, the effect will be explained.

First, standardization of vehicle number photographing size will be explained. Even if it is assumed that there is no calculation error of the vehicle center xo and the calculation error of the vehicle tip y1 at points A and B, there may be errors in the setting of the optical system such as the camera focal length or errors in the photographing position at points A and B. Since it must be assumed that there is an error in the photographed size of about 10% due to the above reasons, the partial images obtained at point A and point B usually do not match. Standardization of the photographic size of a two-dimensional image corresponds to image processing of enlargement and reduction. If this image processing were to be handled by software, it would be impractical because it would require advanced processing techniques such as coordinate transformation, increase the processing capacity, and increase the processing time. In the present invention, instead of transmitting all of the information shown in FIGS. 9 and 10 as a one-dimensional signal at point A, only a standardized number of signals is transmitted as shown in FIG. 12. Then, assuming that the position of (xo, yo) is the same in the image of point A and the image of point B, the partial image of point A in FIG.
It is determined whether the image matches the large partial image of the area of point B shown in FIG. 3 or the small partial image of the area of point B shown in FIG. That is, by applying the one-dimensional digital signals read from the memories 111 and 112 to the D/A converters 113 and 114, analog signals as shown in FIG. 15A, B, and C can be obtained. It is determined whether the patterns of the partial images match based on whether there is a degree of correlation between the analog signals. 15th
Figure A is an analog signal corresponding to the one-dimensional signal of B1 → B2 → B3 in Figure 13, Figure 15 B is an analog signal corresponding to the one-dimensional signal of A1 → A2 → A3 in Figure 12, and Figure 15 C show analog signals corresponding to the one-dimensional signals B1'→B2'→B3' in FIG. 14, respectively. Here, number of pixels × readout clock interval = 2ΔH′ × CL1,
It is assumed that read clocks CL1, CL2, and CL3 are used such that 2ΔH×CL2 and 2ΔH1″×CL3 are constant.

In the above explanation, (xo,
I assumed that the positions of the images in yo) were the same, but
In reality, since there are detection errors and the like, it is necessary to determine whether the image patterns match or do not match due to positional shift. To simplify the explanation, it is assumed that the vehicle number photographing size and the vehicle tip y1 at point A and point B are the same. Figure 16 A and B show the vehicle center line xo
This shows the pattern of position shift when is shifted by Δx. When the partial image at point A as shown in FIG. 16A matches the partial image obtained by shifting the center line xo of the partial image at point B indicated by diagonal lines by Δx as shown in FIG. 16B, A1→ One-dimensional serial signal from A2 to A3 and B1 of the partial image at point B indicated by diagonal lines
By taking an analog correlation with the one-dimensional serial signal of →B2→B3, it is possible to determine whether it is a match or a mismatch. The above also applies to the case of position shift due to the detection error of the vehicle tip y1.

Standardization of vehicle number photographing size and vehicle center line mentioned above
In the positional shift of xo and vehicle tip y1, whether the one-dimensional serial signal at point A and the one-dimensional serial signal at point B match or do not match is determined by the high-speed analog multiplier 115 and comparator 116 in FIG. do.
That is, the first
A one-dimensional serial signal at point A as shown in Figure 7A,
When the high-speed analog multiplier 115 multiplies the one-dimensional serial signal at point B as shown in FIG. 17B given by the A/D converter 114, if both signals match, the signal shown in FIG. As shown by the diagonal line, the analog multiplier output increases. Therefore, by setting a certain threshold level in the comparator 116, it is possible to ensure that both signals match.
Discrepancies can be determined. Note that this cross-correlator 11 does not use the D/A converters 113 and 114, uses a dedicated digital multiplier instead of the analog multiplier 115, and uses a digital comparator instead of the comparator 116. A digital configuration is also possible.

Since the vehicle number image matching device according to the present invention has the configuration and functions as described above, it is possible to determine the vehicle number matching of vehicles passing through point A and point B in real time within one second. For example, the maximum number of samples of partial images sent to point B is 50, the number of pixels of the sample line is 100, and the standardization of the car number shooting size is 10 ways.
Shift the center line xo 10 times and shift the vehicle tip y1 10 times to adjust the clock readout speed.
Assuming 10 MHz, 0.1 μsec x 50 x 100 x 10 x 10 x 10 = 0.5 sec, making it possible to judge in real time within one second. Furthermore, as shown in the embodiments, it is possible to realize a highly practical vehicle number image matching device that can be easily configured using current technology. In addition, even in travel time measurement where the recognition rate is low, it is thought that detection can be made in the same way by using night lighting or by targeting the rear car number of a vehicle that is illuminated at night. Furthermore, the present invention can be expected to be applied not only to traffic control systems but also to parking lot systems and the like.

[Brief explanation of drawings]

Fig. 1 is a block diagram of the vehicle number image matching device according to the present invention, Fig. 2 is a diagram showing image information of a vehicle by an imaging camera, and Fig. 3 is a diagram showing the arrangement of the imaging camera and its monitoring in the case of single lane monitoring. Figure 4 shows the three-dimensional view of the monitoring field in Figure 3, viewed from the side, Figure 5 is the same view from the plane, Figure 6
The figure shows the monitor screen, Fig. 7 shows the car number matching concept flow, Fig. 8 shows the partial image, Fig. 9 shows the two-dimensional image memory of the partial image, and Fig. 10 shows the car number matching concept flow. A diagram showing a one-dimensional transformation of a partial image,
FIG. 11 is a block diagram of the cross-correlator, FIG. 12 is a diagram showing the transmission state of a partial image at point A, FIGS. 13 and 14 are diagrams showing area enlargement and reduction of a partial image at point B, and FIG. Figure 15 A, B, and C are diagrams showing the state of one-dimensional signals at points A and B. Figure 16 A, B
17 is a diagram for explaining the positional shift of partial images at points A and B, and FIG. 17 is a diagram for explaining cross-correlation determination. 1, 6... Imaging camera, 2, 7... Video amplifier, 3, 8... A/D conversion/image memory, 4, 7
...Processing device, 11...Cross correlator.

Claims (1)

[Claims] 1 In a device that verifies the vehicle numbers of the same vehicle passing two points, the center line of the vehicle is determined from the image signal obtained by the imaging camera, and the center line of the vehicle is determined at a certain distance from the front end of the vehicle. A part that cuts out a two-dimensional partial image including a car number image defined by a vehicle height line and vehicle width lines on the left and right sides of the center line, and converts this two-dimensional partial image into a one-dimensional serial signal; By standardizing the field of view mechanics caused by imaging with an imaging camera on the serial signal,
A vehicle number image matching device characterized by comprising a part that aligns the deviation of the center line and the deviation of the vehicle height line with each other and performs cross-correlation on one-dimensional serial signals at two points.