JPH01245400A - Vehicle number recognizing device - Google Patents

Abstract

PURPOSE:To reduce the whole data volume while keeping a number plate feature value by extracting the picked-up image of the front of a vehicle and executing processing such as differentiation, binarization, feature extraction, and compression. CONSTITUTION:The TV-picked-up image signal of the front of the vehicle including the number plate is stored in a frame memory 4 through an A/D converter 2 and a distributor 3 in a vehicle number recognizing device body 20 including a CPU 11, the output of the converter 2 through the distributer 3 is supplied through an input controller 5 and the feature of the vehicle number on the number plate based upon a image density difference is determined by a differentiating circuit 6 and a binarizing circuit 7, a feature extracting circuit 9 and compressed by a compressing circuit 9 and stored in a feature buffer 10. The position of the number plate is estimated from the storage position, corresponding frame data in a memory 4 is binarized and the binary data are stored in a binary image memory 13, so that the whole data volume can be reduced while keeping the number plate feature value. Consequently, the vehicle number of the number plate can be rapidly and surely recognized by respective pattern matching of the circuit 10 and the memory 13.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a vehicle number recognition device for reading a vehicle number, and more particularly to a device capable of measuring the size of a license plate, identifying the type of vehicle, etc.

[Prior Art] Generally, in toll roads, parking lots, etc., it is required to read vehicle numbers by image processing. In this case, the vehicle number is usually recognized by processing the image taken by the left TV camera using an image processing device, but there are problems as follows.

[Problems to be Solved by the Invention] When trying to recognize the vehicle number by detecting the license plate from the image taken by the TV left camera of the front of the vehicle,
An imaging means is required that has a field of view that allows the license plate to be completely captured in the image and has sufficient resolution for vehicle number recognition. When image processing is performed on the obtained information, there is a problem in that the information contains a huge amount of data and cannot be processed quickly.

SUMMARY OF THE INVENTION Therefore, the present invention aims to provide a vehicle number recognition device that can reduce the overall amount of data while maintaining the characteristic amount of the license plate, and can quickly and accurately identify the vehicle number. purpose.

[Means for Solving the Problems] In order to solve the above problems and achieve the objects, the present invention takes the following measures. In other words, there is a TV left camera that images the front of the vehicle including the license plate and sends out the image information as an analog signal, a converter that converts the analog signal input from the TV left camera into a digital multivalued image, and the image generated by this converter. a frame memory for storing a digital multi-value image generated by the converter; a means for selectively retrieving the digital multi-value image generated by the converter or the digital multi-value image stored in the frame memory; A differentiation circuit that performs differentiation on the target image to obtain a differential image; a binarization circuit that binarizes the differential image from this differentiation circuit; The present invention includes a feature extraction circuit that extracts features from an image, a compression circuit that compresses the feature image obtained by the feature extraction circuit, and a feature buffer that stores the compressed image.

[Effect] By taking the above measures, the following effects are achieved.

That is, since the data amount of the target image can be compressed while maintaining the characteristic amount of the license plate, the time required for image processing is significantly shortened, and the plate number can be quickly recognized.

[Embodiment] FIG. 1 is a block diagram showing the configuration of an embodiment of the present invention. The TV left camera is installed in the tollgate gate lane of the H Tollway, takes an image of the front of the vehicle, and outputs image information as an analog signal to the vehicle number recognition device.

The A/D converter 2 receives an analog signal from the TV left camera and converts it into a digital multivalued image. The distributor 3 distributes the digital multivalued image received from the A/D converter 2 to the frame memory 4 side and the front side of the feature extraction circuit. frame memory 4
is for storing the original image, that is, the digital multivalued image. This image is used to try out the image, cut out the plate frame after estimating the plate location, cut out the vehicle number, recognize the vehicle number, etc. The input controller 5 switches the image input to be supplied to the feature extraction circuit group. In other words, the switch is made to the distributor 3 side during imaging, and the switch is made to the frame memory 4 side when retrying. The differentiation circuit 6 differentiates the manually input original image in the horizontal direction. That is, it has a function of determining and outputting the change in density of the pixel of interest with its neighboring pixels. The binarization circuit 7 binarizes the input differential image using a threshold Th set by the CPUI 1. The feature extraction circuit 8 horizontally scans the differential binarized image using a window of a predetermined width, determines the number of "1" pixels within the window, and outputs it as the value at the center of the window. In other words, it extracts the state in which ``dots with large concentrations of eclipse are lined up horizontally'' which is a characteristic of license plates. The compression circuit 9 divides the feature extraction image into unit images of 16 pixels (horizontal) x 1 pixel (vertical), and extracts the pixel having the maximum value from each 11t image as a representative point. Furthermore, this value is binarized using a predetermined threshold value.

Here, the unit image with a value of "1" is determined to have "characteristics". The feature buffer 10 is a memory in which evaluation matrices of each unit image are stored.

The CPU 1 performs overall control of the vehicle number recognition device, input/output control with external equipment, plate frame cutting processing, single character recognition processing, and the like. Note that 12 is a RAM, and 13 is a binary image memory, which stores an image obtained by converting the original image stored in the frame memory 4 into a binary image.

In FIG. 1, the part excluding the TV left camera, that is, the part 20 surrounded by a broken line, is the main body of the vehicle number recognition device.

Next, the operation of the apparatus configured as shown in FIG. 1 will be explained with reference to the flowchart shown in FIG. TS2.

An analog signal representing an image of the front of the vehicle captured by the TV left camera is converted into a digital multivalued image by the A/D distributor 2. This digital multivalued image is sent to the front side of the feature extraction circuit by the distributor 3, and is also stored in the frame memory 4 as an original image for retry vehicle number reading. The digital multivalued image outputted from the distributor 3 to the feature extraction circuit fused copper is manually inputted to the differentiation circuit 6 via the manual controller 5. Hereinafter, the processing is sequentially performed according to each step shown in the flowchart of FIG.

[Step]] The digital multivalued image is differentiated in the horizontal direction. In other words, the density difference between adjacent pixels is calculated,
Its absolute value is output.

[Step 2] The differential image is binarized by a binarization circuit 7 using a predetermined threshold Th. At this time, a point having a density difference larger than the threshold value is set as "1", and the background is set as "0".

Note that the binarization circuit 7 has three types of threshold values Th, -Th3, and retry is performed by converting these binarization threshold values. In the license plate part of the image, there is a large density difference between the characters and other parts, so when differentiation is applied, many points with high values appear. Therefore, the number of points that become "1" increases due to binarization.

Estimation of the location of the plate is performed by paying attention to the way the points that become "1" are gathered.

[Steps 3 and 4] The feature extraction circuit 8 extracts the neighborhood (
The number of "1"s existing in the horizontal direction) is taken as information about that point. Next, the characteristic image is divided into unit images of 166 pixels (horizontal) x 1 pixel in this embodiment, and the maximum value is extracted from each qt image as a representative point. In other words, it is compressed. Furthermore, this compression is performed using the J method that extracts the maximum value, so the feature values are not impaired in the plate portion.

[Step 5] The compressed feature image is binarized using a predetermined threshold Th2 and stored in the feature buffer 10. Among the data stored in the feature buffer 10, a point that is "1" means "a portion in which there are many high-density points nearby." In other words, it shows that the +17 potential that is part of the plate is high. Up to this point, extraction has focused on the horizontal features of the plate.

[Step 6] Next, the position of the plate is estimated by paying attention to the vertical alignment of the extracted portions. This process is performed by CPU II. Here, the features in the vertical direction of the plate in the feature binary image will be explained. The reason why many feature points appear in the license plate area in the horizontal feature extraction described above is because there are characters and other areas within the plate that have a significant difference in density. However, the inside of the license plate There are also parts that do not appear in the horizontal features.

Figure 3 is a diagram showing an example of a license plate, and Figure 4 shows an example of a license plate.
The figure is a concentration distribution diagram showing the concentration distribution in the horizontal direction of the serial number (capital letter) part, the lowercase letter part, and the part in between in the same plate. As shown in the figure, the uppercase line part L
There is a drastic change in density between 3 and the lowercase line part L1,
Since there are no characters in the line portion L2 in the middle, there is almost no change in density. Therefore, even within the plate, no horizontal feature appears in the area between the uppercase and lowercase letters. Therefore, if the images stored in the feature buffer 10 are searched vertically, "i" is found in the plate part.
A pattern should appear.

FIG. 5 is a diagram showing the internal state of the feature buffer 10. The black portion in the figure is the portion where features appear in the horizontal direction, and is the length of the J゛-modulo a unit image in the horizontal direction. CjUll based on the above viewpoints
The estimation of the plate location is performed by vertically evaluating the contents of the feature buffer 10 and detecting the "i" pattern. In addition, at that time, it is necessary to consider whether the vertical size of the "i" pattern is appropriate for a license plate, so the range of the size of the detected "i" pattern is necessarily determined. Since the position where the "i" pattern is detected can be regarded as the position where the habo plate exists, the estimation is completed. Next, the CPU II
” The plate frame is cut out based on the position where the pattern is detected.

FIG. 6 is a diagram showing a search range when cutting out a plate.

Since the "i" pattern should accurately indicate the length of the plate in the vertical direction, the vertical search range may be from the top to the bottom of the "i" pattern.

The search range in the horizontal direction needs to be about twice the width of the plate, taking into consideration the case where the extracted "i" pattern exists at the left or right end of the plate.

In order to cut out the plate frame, the CPU II first binarizes the original image stored in the frame memory 4 within the above-mentioned range and stores it in the binary image memory 13. The threshold value used in the binarization is determined by "Otsu's method" for the range shown in FIG. 7 (the part surrounded by the dotted line). "
If the "i" pattern exists on the license plate, most of the range shown in FIG. 7 would be occupied by the density corresponding to the letters and the iH degree corresponding to others. Therefore, it is possible to find the optimal binarization threshold using "Otsu's method" or the like.

The uppercase portion within the binarized estimated region, that is, the image of the rectangular range surrounded by kI * k2 + J+ 1 in Fig. 6, is projected in the direction of arrow A in Fig. 6, and the area is projected in the direction of arrow A in Fig. 7. is indicated by the dotted line.

FIG. 8 is a diagram showing the imaging results. The projection result 25 is binarized using a threshold Th3 and becomes a binary one-dimensional pattern 26. The threshold value Th3 is a value that allows one character line to be detected, and is relatively easy to determine.

Next, note the length of the part corresponding to a character in this one-dimensional pattern, the number of one character in length corresponding to other parts, etc., and match it with the pattern shown in Figure 9, for example, to create a plate. Detect the width of

It is possible to detect the plate frame by the above method.

[Step 8] As shown in FIGS. 10 and 11, the vehicle number contained within the license plate frame is in a fixed position. Therefore, if the plate frame can be detected, it is possible to extract the vehicle number using the Y-axis coordinate Vlly2 in FIG. 6, which is information regarding the characters already determined. This is done by the CPU 11 by evaluating the binary image memory used for cutting out the plate frame.

[Step 9] The CPU II performs pattern matching on the extracted vehicle number to recognize the characters. Since the characters on the license plate are standardized numbers, it is relatively easy to pattern match the pre-stored pattern of numbers from rOJ to "9" with the cut-out vehicle number.

[Effects of the Invention] According to the present invention, since the target image is compressed while preserving the feature amount of the license plate, the amount of data to be processed is significantly reduced, and the vehicle number can be recognized quickly and accurately. It is possible to provide a vehicle number recognition device that can.

[Brief explanation of the drawing]

1 to 10 are diagrams showing an embodiment of the present invention,
Fig. 1 is a block diagram showing the configuration of the device, Fig. 2 is a flow diagram for explaining the operation of the device, Fig. 3 is a diagram showing an example of a license plate, and Fig. 4 is a diagram of the license plate shown in Fig. 3. Horizontal concentration distribution map, Figure 5 is a schematic diagram showing the internal state of the feature buffer, Figure 6 is a diagram showing the plate search range, and Figure 7 is a diagram showing the area of interest for determining the binarization threshold. , Fig. 8 is a projection result diagram showing the vertical density distribution of the plate frame detection range, Fig. 9 is a diagram showing the matching pattern for plate frame detection, and Figs. 10 and 11 are projection results for the license plate frame. It is a figure which shows the example of a vehicle number position. 1...TV left camera 2...A/D converter, 3...
Distributor, 4... Frame memory, 5... Human controller, 6... Differentiating circuit, 7... Binarization circuit, 8...
...Feature extraction circuit, 9...Compression circuit, 10...Feature buffer, 11...CPU, 12...RAM, 13
...Binary image memory, 20...Vehicle number recognition device main body. Applicant's Representative Patent Attorney Takehiko Suzue Figure 3, Figure 4, Figure 6, Figure 7, Figure 8, Figure 10

Claims (1)

[Claims] A TV camera that images the front of the vehicle including the license plate and sends out the image information as an analog signal, a converter that converts the analog signal input from the TV camera into a digital multi-level image, and a digital image generated by this converter. a frame memory for storing a multivalued image; a means for selectively retrieving the digital multivalued image generated by the converter or the digital multivalued image stored in the frame memory; and a target image retrieved by the means. A differentiation circuit that performs differentiation to obtain a differential value image, a binarization circuit that binarizes the differential value image from this differentiation circuit, and features from the image binarized by this binarization circuit. A vehicle number recognition system comprising: a feature extraction circuit for extracting a feature image; a compression circuit for compressing a feature image obtained by the feature extraction circuit; and a feature buffer for storing the compressed image. Device.