JPH0883393A - Vehicle speed measuring device - Google Patents

Abstract

(57) [Summary] [Purpose] To add a speed measurement function to a vehicle number recognition device for highly accurate measurement of the speed of a passing vehicle in order to recognize a license plate. [Structure] According to the passage timing of a vehicle, two images are input and a serial number extraction unit that cuts out a serial number of a plate using one image, and a grayscale image of a rectangular area including the serial number is registered as a template. Template registration unit, a correlation calculation processing unit that executes a correlation calculation process between the registered template and the second image, and a moving pixel between the coincidence coordinate obtained by the correlation process and the Y coordinate when the template is cut out. It has a speed calculator that calculates the speed from the number. [Effect] By registering the grayscale image of the serial number as a template and executing the grayscale pattern matching process between this image and another image, one highly accurate detection can be performed, and as a result, the speed calculation precision is also extremely high. Therefore, it is possible to adapt to various traffic control systems.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a vehicle speed measuring device having a function of measuring the traveling speed of a vehicle.

[0002]

2. Description of the Related Art Various vehicle number recognition devices for vehicles traveling on roads have been developed, but usually, there is a device for recognizing a serial number of a license plate or all character information. Such a vehicle number recognition system, as shown in FIG.
Although the vehicle 45 is installed by the TV camera 46, it can be used, for example, as an overspeed warning device as well as for grasping a traffic jam on the road if the vehicle has a passing speed measuring function. As a conventional speed measurement method, there is a method of extracting the same feature point from two images and calculating the speed from the number of moving pixels at that point, as in Japanese Patent Laid-Open No. 5-266394. As disclosed in Japanese Patent Laid-Open No. 6-28596, "Image processing type overspeed warning device", there is a method in which the position of a license plate is obtained from two images and the speed is calculated from the number of moving pixels at that position.

[0003]

In the above-mentioned conventional technique, since the characteristic points such as the central coordinates of the characters (serial number) of the license plate and the position of the license plate are obtained from the two input images, the plate detection and the characters are detected. Since it is necessary to execute the extraction, it takes a long processing time, and since only one feature point is checked, there is a problem that pixel shift easily occurs.

An object of the present invention is to provide a vehicle number recognition device capable of measuring the speed of a vehicle passing at high speed and with high accuracy.

[0005]

The above-mentioned object is to capture two images from a vehicle passage timing, use one of the images to cut out a specific area of a license plate, for example, a serial number, and include the serial number. The rectangular area of the input image is registered as a template image, the grayscale image matching process is executed between this template image and the other input image, and the pixels at the most matching position and the position of the image registered in the template Number of moving pixels from the number,
The speed is calculated using this value.

[0006]

The amount of movement of the gray-scale image of the plate serial number on the two images is measured by using the gray-scale pattern matching processing. By such processing, it is not necessary to perform plate detection, character extraction processing, or the like for each image, so that speed measurement can be performed at high speed, and since it is pattern matching of grayscale images,
This is a planar alignment and, in principle, subpixel (1
It is possible to obtain the number of moving pixels with an accuracy of (/ 10 pixel unit).

[0007]

An embodiment of the present invention will be described with reference to the drawings.
FIG. 1 shows a flow of speed measurement, and the operation will be described in sequence.

(1) The first image is input from the timing of vehicle detection (step 1), and then the second image is input (step 2). The difference in capture time between the first and second images is 1/60 when processing a field image.
Seconds, 1/30 seconds when processing frame images.

(2) A serial number is extracted using the input first image (step 3). This extraction method is, for example,
The license plate area (specific area) is extracted, and binarization processing is executed only for that area, and only the objects in which the size, character spacing, etc. of each object match the serial number are obtained. Obtained by extracting.

(3) The grayscale image of the input image corresponding to the rectangular area including the extracted serial number is registered as a template (step 4). The minimum Y coordinate of the rectangular area including the serial number at this time is stored as Y1.

(4) The search range of the registered template is determined (step 5). The range of the matching process is small in the horizontal direction because the vehicle moves from the top to the bottom of the screen. For this reason, the horizontal direction is slightly larger than the registered template, and the vertical direction is searched downward from the Y coordinate Y1 of the template for a width corresponding to the number of moving pixels of the upper limit value of the measured speed.

(5) The pattern matching process of the grayscale image is executed between the registered template and the image input to the second sheet. Here, an example using the normalized correlation processing is shown. Here, the normalized correlation processing will be briefly described. The normalized correlation processing is to execute an operation as shown in Expression 1, and a certain point (u, v) is applied to the processing target image f by using the grayscale template T. The similarity r (u, v) of is calculated. This similarity is closer to a value of 1 as the template is more similar.

[0013]

[Equation 1]

Where f (u + i, v + j) is (u, v)
The density value of the target image near the point, T (i, j) is the density value of the (i, j) point of the template image, and p and q are the x and y sizes of the template image, respectively. Since data regarding only the template 31 (FIG. 5) that does not depend on the target image in Expression 1 can be calculated in advance, the values of Expression 2 and Expression 3 can be calculated before the correlation calculation 32, 33 (FIG. 5). Therefore, in the correlation calculation between the template data 31 and the correlation target image 30 (FIG. 5),
Expression 4, Expression 5, and Expression 6 are calculated in real time, and the similarity of Expression 1 is finally obtained using the respective calculation results.

[0015]

[Equation 2]

[0016]

(Equation 3)

[0017]

[Equation 4]

[0018]

(Equation 5)

[0019]

(Equation 6)

(6) Such a calculation is performed over the entire search area to obtain the point at the coordinates (x, y) with the highest degree of similarity. That is, (u, v) of the similarity r (u, v)
Is moved only in the search area, and the (x, y) coordinates of r (x, y) showing the maximum similarity among the respective similarities become the obtained coordinates. The y coordinate at this time is set to Y2 (step 7). (7) The moving distance is calculated from the pixel at the coordinate Y1 at the time of template registration and the coincident coordinate Y2 obtained by the correlation calculation, and the speed is calculated (step 8).

FIG. 2 shows the concept of speed calculation. Camera 10
Is installed downward at the position of height H. It is assumed that the plate position is obtained as 9 in the first image and 9'is obtained in the second image. At this time, the position L1 of the first plate is obtained by the equation 7, and the position L2 of the second plate is obtained by the equation 8.

[0022]

(Equation 7)

[0023]

[Equation 8]

Here, h indicates the height of the plate, which is about 50 cm for passenger cars. θ1 is the angle with the camera corresponding to the position of the upper part of the plate 9 in the first image, and θ2 is the angle of the plate 9 ′ similarly,
It is calculated using Equation 10.

[0025]

[Equation 9]

[0026]

[Equation 10]

Δφ in Equations 9 and 10 is the vertical field of view φ of the camera.
Is a value obtained by dividing by the Y width when stored in the image memory for image processing, and corresponds to the viewing angle per pixel. ρ indicates the frontmost angle of the camera. Y1 is the top Y coordinate of the serial number of the plate obtained from the first image, and Y2 is the top Y coordinate of the serial number obtained from the second image.

By carrying out such a calculation, the speed is calculated from the difference between L1 and L2 and the difference in the capturing time of the first and second images.

FIG. 3 shows a hardware configuration of the present invention, which is an A / D for digitizing a video signal of the camera 10.
An image is input to the image memory via the converter 11. This image memory is a grayscale image memory of about 8 bits.
The first image is input to 12 image memories # 1 and the second image is input to 13 image memories # 2. A serial number area cutout unit 14 cuts out an area image having a serial number from the first image. Then, the template registration unit 15 registers the image of the area as a template.

Correlation processing is performed between the registered template and the image of the 13th image memory # 2 in which the second image is stored by using the correlation calculation unit 16, and the Y coordinate of the template,
The speed calculation unit 17 calculates the speed from the difference between the Y coordinates of the correlation coincidence coordinates and the like.

Description will be made with reference to a specific vehicle drawing. FIG.
When the first image is obtained as shown by g1, the sequence number extraction means 20 extracts the sequence number from this image. A grayscale image 22 having a serial number of a rectangular area including this serial number is cut out at 21. The top Y coordinate of the serial number at this time is Y1. For the second image g2, the search area is set so that the starting point Y coordinate is Y1 and the ending point Y coordinate is downward with a predetermined width. The search width of X is set to be a little wider than the width of the serial number because the horizontal movement is small. Correlation calculation with the template is executed for this search area, and the maximum matching coordinate is detected at 23. With the detected Y coordinate as Y2, the speed is calculated at 24 from the values of Y1 and Y2.

FIG. 5 shows the hardware structure of the normalized correlation processing, which is a structure for calculating the values of the expressions 2 to 6 in the expression 1 in parallel. By using such hardware, the correlation calculation can be executed at high speed.

As described above, the number of moving pixels of the vehicle can be accurately detected by executing the density correlation calculation using the grayscale image of the serial number as a template, so that the passing speed of the vehicle can be accurately measured. Is.

In the above embodiment, the input image is used as it is, but in order to shorten the processing time, the input image may be subjected to the correlation calculation with the image reduced to ½ or ¼. It is possible. At this time, an affine transformation function may be provided as an image processing function.

In the above embodiment, the grayscale image including the serial number is used as the template. However, the grayscale image of the rectangular area including the images of the Land Transport Office code and the vehicle type code, the grayscale image of the entire plate, etc. should be registered in the template. However, the same effect can be obtained. However, there is a problem that the processing time increases as the size of the template increases. Furthermore,
Although the specific area of the plate is cut out using the input first image, even if the specific area is cut out using the second image and this image is correlated with the first image. The same effect is obtained.

Further, in the above-mentioned conventional example, the method of finding the maximum matching coordinate is used.
It is also possible to perform the sub-pixel positioning using the similarity of (pixel), and it is possible to significantly improve the speed measurement accuracy as compared with the conventional method.

(Other Embodiments) In the above embodiments, it is not possible to measure the speed of a vehicle traveling at an ultrahigh speed. That is, when the time for passing through the field of view of the television camera becomes high, even if the image is captured at the video rate or the image 38 of the first license plate is captured as shown in FIG. 6,
In the case of the second image, a case where an image in which the plate is not in the field of view is taken in as in 39 occurs, which makes the velocity measurement impossible.

Therefore, two TV cameras are arranged vertically in FIG.
Installed adjacent to each other as shown in, to enlarge the apparent vertical field of view. Then, as shown in FIG. 8, the two images C1 (1) and C1 (2) are input from the camera C1 from the timing when the vehicle passes through the field of view of the camera C1.
Simultaneously with the timing of capturing the first image, the camera C2 also captures one image C2 (2). Normally, serial number extraction of the license plate is performed using C1 (1) above, and velocity measurement is performed by correlation processing using the plate image and the image of C1 (2), but an image with high correlation is found. If not,
It is determined that the image has passed through the field of view of the camera C1, and the plate image of C (1) and the image of C2 (2) are subjected to correlation calculation to perform speed measurement. By doing so, it is possible to measure the speed of the passing vehicle at an extremely high speed. In this case, since it is necessary to synchronize the cameras C1 and C2 with each other, C1 and C2 are cameras of an external synchronization system.

Since the viewing angle of the camera is wide in the horizontal direction, it is somewhat effective to install the camera laterally to widen the viewing angle in the traveling direction of the vehicle, but conversely the running width cannot be viewed wide. Therefore, there is a problem that the probability of catching the plate on the screen is reduced.

[0040]

As described above, two images are sequentially input from the timing when the vehicle passes through the field of view of the camera, the serial number of the plate is extracted using one image, and the grayscale image of the rectangular area including this serial number is used as a template. Registered as this image and 2
By performing the correlation calculation between the images of the first sheet and measuring the moving amount of the plate, it is possible to measure the passing speed with high accuracy.
It is also possible to install two cameras in the vertical direction and measure the speed of a high-speed vehicle.

[Brief description of drawings]

FIG. 1 is a diagram showing an operation of a passing vehicle speed measurement process of the invention.

FIG. 2 is a diagram illustrating speed calculation.

FIG. 3 is a diagram illustrating a concept of a hardware configuration for speed measurement processing.

FIG. 4 is a diagram illustrating a specific example of speed measurement.

FIG. 5 is a diagram showing hardware of a correlation calculation process.

FIG. 6 is a diagram showing a passing state of an ultra-high speed vehicle.

FIG. 7 is an installation diagram when measuring the speed of an ultra-high-speed vehicle using two cameras.

FIG. 8 is a diagram illustrating a concept of speed measurement processing in the case of two cameras.

FIG. 9 is a conceptual diagram of camera installation of a vehicle number recognition device.

[Explanation of symbols]

10 ... TV camera, 12, 13 ... Image memory, 14 ...
Serial number area image cutout unit, 15 ... Template registration unit, 16 ... Correlation calculation unit, 17 ... Speed calculation unit.

Front page continuation (72) Inventor Yoshiyuki Sato 5-2-1 Omika-cho, Hitachi-shi, Ibaraki Hitachi Ltd. Omika factory

Claims (7)

[Claims] 1. A photographing means for photographing a moving vehicle with a television camera installed above a road, and at least two images of the image signal of this photographing means are sequentially fetched and stored at predetermined time intervals from the vehicle detection timing. Image memory, image cutout means for extracting a grayscale image of a specific area of the license plate from one of the two images in the stored image memory, and template registration means for registering the cutout image as a template. And the position of this registered template and 2 stored in the image memory
A vehicle speed measuring device comprising a speed calculating means for calculating a speed based on a position of an image having the highest degree of similarity in the other image of the images. 2. The vehicle speed measuring device according to claim 1, wherein the specific area of the plate is a grayscale image including a serial number. 3. The vehicle speed measuring device according to claim 1, wherein the specific area of the plate is a grayscale image including a Land Transport Office code and a vehicle type code. 4. The vehicle speed measuring device according to claim 1, wherein a specific area of the plate is a grayscale image of the entire license plate. 5. The vehicle speed measuring device according to claim 1, wherein the correlation calculation processing unit has a function of executing a normalized correlation process of the grayscale image. 6. The speed calculation unit according to claim 1, wherein the speed calculation unit calculates a distance to the camera position obtained from the Y coordinate of the specific area at the time of template registration, and calculates a distance to the camera position obtained from the Y coordinate obtained by the correlation processing. A vehicle speed measuring device, which calculates and calculates a speed from a difference between these distances and a time difference between capturing two images. 7. A photographing means for photographing a running vehicle by vertically arranging two television cameras A and B above the road, and a video signal of the television camera A of the photographing means, two images from a vehicle detection timing. Of the video signal of the TV camera B and the timing of inputting the video signal of the TV camera B at the same time when the TV camera A inputs the second video.
An image memory that captures and stores images, and an image cropping unit that extracts a grayscale image of a specific area of the plate from the first image of the two images in the image memory that stores the image of the TV camera A, and the cropped image A template registration unit that registers as a template, a correlation calculation processing unit that performs a correlation calculation between the template and the specified grayscale image,
Correlation calculation processing is performed between the registered template and the second image of the two images in the image memory in which the image of the TV camera A is stored. If the maximum correlation value is larger than the predetermined value, the matching coordinates Then, the speed is calculated from the number of moving pixels of the Y coordinate of the registered specific area, and when the maximum correlation value is smaller than a predetermined value, the registered template image and the image of the image memory in which the image of the television camera B is stored. And the correlation processing is executed between the
A vehicle speed measuring device comprising: a speed calculation unit that calculates a speed based on the coordinate and the Y coordinate of the maximum matching coordinate.