JPH11175655A - Vehicle number reader - Google Patents

Abstract

PROBLEM TO BE SOLVED: To prevent the character recognition ratio of a number plate from being reduced by preventing the generation of uneven brightness on the number plate. SOLUTION: In the device for detecting an image of a number plate from a video signal outputted from a video camera and reading out characters from the detected number plate image, the upper part P or the whole of the number plate is divided by plural horizontal lines as a signal binarizing threshold for segmenting or recognizing each character from the number plate image. A threshold is found out from the lightness of each divided area, and thresholds corresponding to respective areas are adopted.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an apparatus for photographing a running automobile with a camera and recognizing license plate numbers and characters from a video signal.

[0002]

2. Description of the Related Art In a "license plate position detecting device" (Japanese Patent Application Laid-Open No. 62-276696) previously submitted by the present applicant, each time a video signal for one frame is stored in a memory from a camera, the video signal is output. By binarizing the signal and comparing the arrangement of the white and black dots with an algorithm, it is recognized whether or not the data includes data relating to a license plate. Further, in the "method of detecting a display unit for identifying an article" (Japanese Patent Laid-Open No. 2-224083), a distribution mechanism recognizes a delivery destination code described on a label attached to a product. In both of these inventions, instead of recognizing the license plate or attached label after confirming that the car or the product has entered the shooting area of the camera, the license plate or attached label is directly read from the screen that is constantly shooting. It has a feature in recognition.

An example of the former application is, for example, an introductory management system. The license plate of the car to be entered is read, the number is checked against the registered data, and if the car is permitted, the entrance gate is automatically opened. In this system, a video signal for one frame is binarized by an appropriate threshold value, and if the binarized data includes data relating to a license plate, a character portion is cut out from the data of the license plate portion. , And recognizes characters from the character part.

[0004]

By the way, as shown in FIG. 1, a license plate has a car number written on a lower part Q thereof.
In the upper part P, a land transportation branch office name and a vehicle type number are written.
Depending on the state of attachment of the license plate, a shadow due to a bumper or the like may be formed on the upper portion P of the license plate, causing uneven brightness on the license plate. In the above-described introductory management system, when binarizing an image for a license plate portion, conventionally, a single threshold is used, so that a shaded character is small and cannot be cut out or recognized. In some cases, the reading rate decreased.

Accordingly, an object of the present invention is to provide a device capable of accurately reading out characters on a license plate even when a shadow is produced on the license plate.

[0006]

The brightness unevenness generated on the license plate changes in the vertical direction of the license plate but hardly in the horizontal direction as shown in FIG. Therefore, as shown in FIG.
Is divided by a plurality of horizontal lines, a threshold value is obtained from the brightness detected for each of the divided areas, and a corresponding threshold value is adopted for the data of each of those areas.

When FIG. 2 shows data stored in a memory, the data is stored in the memory (second storage means) for each one or a plurality of horizontal scanning lines. The data is read, a threshold value is obtained from the brightness of the read data, and the data is binarized by the corresponding threshold value.

[0008]

FIG. 3 is a control block diagram showing one embodiment of the present invention. Reference numeral 1 denotes a camera having a CCD image sensor, which outputs a non-interlaced one-frame image at a period of 1/30 second. Reference numeral 2 denotes a lens mounted on the camera 1 and has an aperture mechanism inside. 3
Is an A / D converter for A / D converting an analog video signal from the camera 1. Reference numeral 4 denotes a frame memory that stores the A / D converted video signal for one frame.
As shown, it is composed of two systems of frame memories.

Reference numeral 5 denotes a plate detection unit for detecting whether data relating to a license plate image is included from the data stored in the frame memory 4, and can use the above-mentioned known technique. Reference numeral 6 denotes a CPU, which performs “average photometry processing” for changing the aperture value and the shutter speed based on the amount of light incident on the camera 1 when the data for one frame does not include license plate image data. ,
When a license plate image is detected, “plate photometry processing” for changing the aperture value and shutter speed from the data for the license plate is performed. The data obtained by the photometry is sent to the camera 1 aperture control circuit 7 for driving the aperture mechanism of the lens 2 and the shutter control circuit 8 for determining the shutter speed of the camera 1.

The number and characters of the license plate recognized by the CPU 6 are supplied to, for example, a host computer 11 and compared with a vehicle number registered in advance in a database 11a, so that the vehicle is approved for the vehicle. If there is, a signal for opening the entrance gate is output. The recognized vehicle number is displayed on the monitor 12.

The control operation of this device is shown in the flowchart of FIG. As described above, a one-frame video signal is output from the camera 1 at a period of 1/30 second. When the one-frame video signal is stored in one frame memory 4 (I) in step S1, the next Output to 1
The video signal of the frame is stored in the other frame memory (II). Thus, the video signal is stored alternately in the frame memories 4 of two systems.

When a video signal is stored in the frame memory 4 (I), the video signal of the previous frame stored in the other frame memory 4 (II) is compared with the video signal of one frame in step S2. The plate detection circuit 5 checks whether a plate is included. Step S
At 3, the presence or absence of a license plate is determined. If the license plate is not included, the process ends. If the license plate is included, the process proceeds to step S4, where the video signal of the upper part P of the license plate is displayed as shown in FIG. As shown, it is stored in the array image aij. In step S5, a threshold value for each horizontal scanning line is determined for this image. The detailed operation is shown in a subroutine of FIG.

In step S51, the value of i is initialized to 0. In step S52, X is applied to the array image aij.
, And data a ₀₀ , a ₀₁ ... For one scanning line.
a _0j-1 is read. In step S53, the maximum value and the minimum value and is detected when the difference between the maximum value and the minimum value as the contrast in step S54 is calculated, the contrast is stored in an array C ₁ shown in FIG. 2 . In step S55, the value of i is incremented to scan the horizontal scanning line in the Y direction.
In step S56, it is determined from the value of i whether all the horizontal scanning lines have been scanned in the Y direction (that is, the entire image array aij has been read out). The data reading of the line is repeated.

When the scanning in the Y direction is completed, the process proceeds from step S56 to step S57, and the minimum value is detected from the array Ci. The value obtained by multiplying this value by the k value (k = 1.3) is set as the minimum value again. This value is the contrast of the background portion on the plate and is defined as the signal level.

In step S58, the value of i is initialized again, and in step S59, the array Ci, which is the contrast, is initialized.
It is determined whether or not (i at this time is 0) is lower than the signal level. If so, in step S61, 0 is set as a threshold value for the data read in the horizontal scanning of the first row. Otherwise, in step S60, an average value of the maximum value and the minimum value obtained in the scanning of the first row is set as the threshold value. Step S
In 62, the threshold value obtained in step S60 or step S61 is stored in the above-described array Ci. In step S63, the value of i is incremented. In step S64, it is determined whether or not the scanning in the Y direction has been completed based on the value of i. The threshold is determined. When all the threshold values are stored in the arrays C _{0 to} C _i-1 in this manner, the process returns to the flowchart of FIG.
Here, the threshold value is set for each scanning line, but one threshold value may be set for every several lines based on the size of the image on the memory.

In step S6, each image is binarized with the corresponding threshold value Ci for each horizontal scanning line with respect to the image of the plate upper part P. In step S7, individual characters are cut out from the binarized image.
The character cut out in step 8 is recognized by a known character recognition unit.

From step S11, the lower plate Q
Character recognition is performed on (an area with a car number) according to a conventional method. That is, in step S11, the image of the plate lower portion Q is binarized by a predetermined threshold value (the threshold value used when the license plate is found in step S2 or a preset fixed threshold value), In step S12, characters are cut out from the binarized image, and in step S13, the cut-out characters are recognized.

Thereafter, the process returns to step S1, and character recognition is performed on the signal stored in the frame memory 4 (I).

In the present embodiment, the lower part of the license plate Q where shadows are less likely to be generated, in order to speed up the processing,
Although one type of threshold is used, if there is enough processing speed, as shown in the flowchart of FIG. 6 according to the second embodiment of the present invention, in step S4 ', the license plate is added to the image array aij. Is stored, and the same processing as performed in the upper part P is applied to the lower part of the license plate Q, so that the character extraction and the character recognition for the lower part of the license plate Q can be performed more reliably.

FIG. 7 is a flowchart showing the third embodiment. Steps S11 to S13 are:
This is the same as steps S1 to S3 in FIG.
In the next step S14, a binary fixed value is used for the image of the upper part P of the license plate (the threshold used when the license plate is found in step S12 or a fixed threshold set in advance). In step S15, characters are cut out from the binarized image.

In step S16, the cut-out characters are stored in the image array aij. In step S17, a threshold value for each horizontal scanning line is determined for the image using the subroutine of FIG. Then, in step S18, the original data of the character portion that has not been binarized is binarized using the threshold value, and in step S19, a character is recognized from the binarized image. In steps S21 to S23, similar to steps S11 to S13 in FIG. 3, a known character recognition method is adopted for the plate lower part Q.

In the control shown in FIG. 7, the character cutout is performed at a fixed threshold, so that the success rate of the character cutout may slightly decrease. However, the threshold is set for each horizontal scanning line only for the cutout character area. Since the value is determined, the processing time can be reduced as compared with the control in FIG. 4 (character cutting and character recognition at the top of the plate are performed using a threshold value for each horizontal scanning line).

In the present embodiment, one kind of threshold value is adopted for the lower part of the license plate Q in which a shadow is unlikely to occur, as in the flow of FIG.
If there is enough processing speed, as shown in the flowchart of FIG. 8 according to the fourth embodiment of the present invention, in step S14 ′, all the images of the license plate are stored in the image If the same control as that performed on the license plate upper part P is applied to the plate lower part Q, character cutout and character recognition for the license plate lower part Q can be performed more reliably.

The operation performed on the license plate upper part P (steps S14 to S19) in the flow of FIG. 7 may be replaced with the operation performed on the license plate lower part Q (steps S11 to S13) in FIG. . As described above, various processing modes are conceivable. In any case, it is essential to set a threshold value for each horizontal scanning line.

[0025]

As described above, according to the present invention, the upper part or the whole of the license plate is provided with a plurality of horizontal lines as a threshold for signal binarization for character extraction or character recognition from the license plate image. In the above, a threshold value is obtained from the brightness of each of the divided regions, and a corresponding threshold value is employed for each of the regions. Therefore, even if there is uneven brightness in the vertical direction on the license plate, Character recognition can be performed accurately.

[Brief description of the drawings]

FIG. 1 is a view showing the form of a license plate.

FIG. 2 is a diagram showing data on a license plate memory.

FIG. 3 is a control block diagram of the vehicle number reading device of the present invention.

FIG. 4 is a control block diagram showing a first embodiment of the above device.

FIG. 5 is a flowchart showing a subroutine in the flow of FIG. 4;

FIG. 6 is a control block diagram showing a second embodiment of the above device.

FIG. 7 is a control block diagram showing a third embodiment of the above device.

FIG. 8 is a control block diagram showing a fourth embodiment in the above device.

[Explanation of symbols]

 Reference Signs List 1 camera 2 lens 3 A / D converter 4 frame memory 5 plate detection circuit 6 CPU 6a ROM 7 aperture control circuit 8 shutter control circuit

Claims (2)

[Claims] An apparatus for detecting an image of a license plate from a video signal of a video camera and reading a character from the detected license plate image, wherein a signal for binarizing a character from the license plate image or performing character recognition is used. , The upper part or the whole of the license plate is divided by a plurality of horizontal lines, a threshold value is obtained from the brightness of each of the divided regions, and a corresponding threshold value is adopted for each of the regions. A vehicle number reading device, characterized in that: 2. A / D conversion means for A / D converting a video signal output from a video camera, a first memory for storing the A / D converted video signal, and a first memory for storing the A / D converted video signal. Plate detection means for detecting data relating to the image of the license plate from the data of the one frame, second storage means for storing the detected license plate as two-dimensional array data, and second storage means. Binarizing means for reading data for one or a plurality of horizontal scanning lines, obtaining a threshold value from the brightness of the read data, and binarizing the data with a corresponding threshold value; And a character recognizing means for performing character segmentation and character recognition based on the digitized data.