TWI603268B - Image processing system and method for license plate recognition - Google Patents

Description

Image processing system and method for license plate recognition

The invention relates to a license plate recognition method, in particular to an image processing system and method for feature coded license plate recognition.

License plate recognition (LPR) is an important part of modern intelligent transportation system. Its research has been carried out for a long time. Through the license plate recognition system, traffic flow control, vehicle positioning, highway overspeed automation supervision, Traffic violations, highway toll collection systems, and parking lot management; help maintain traffic safety and urban security, and save labor costs for law enforcement agencies.

Figure 1 is a flow chart of the conventional license plate recognition. The main reason is that after receiving the image, the brightness and direction of the image should be corrected first because of the natural environment illumination and the shooting angle. After the correction, the position of the character on the license plate is removed. In order to perform character segmentation, the image recognition method is used to perform character recognition, thereby identifying the character of the license plate. At present, most of the identification techniques are neural network learning or sample matching. However, the neural network Road learning requires a lot of pre-training. The more the training samples, the higher the recognition rate will be, and the template matching is to establish a sample library. The normalization of the character size becomes very important during identification, which will lead to the image recognition system. The required storage space is too large, and the processing of image recognition becomes more and more complicated. It does not meet the need for the existing license plate recognition system to be miniaturized and to quickly and correctly identify license plate characters. Therefore, there is still a need for improvement.

Therefore, the object of the present invention is to provide an image processing system and method for license plate recognition, which uses a code table to perform character recognition without storing a standard template in advance, thereby reducing computational complexity and storage space for identification.

According to a feature of the present invention, the present invention provides an image processing system for recognizing a character of a license plate. The system includes: a receiving module for receiving an image of the license plate; and a character segmentation module for Dividing the image of the license plate into a plurality of character regions, each character region includes a plurality of pixels to form a character; and a character recognition module for recognizing the character in each character region The character recognition module first performs a directional feature extraction for each character region, and then divides each character region after the directional feature extraction into a plurality of blocks and determines one of each block. The feature values are combined with the feature values to identify the character.

The simplified conversion system first divides each character region into a plurality of pixel regions, and then compares a pixel point distribution in each pixel region with a plurality of preset patterns, and compares each pixel according to the comparison result. The area is classified, and the classification result replaces the pixel point distribution with the corresponding pattern.

Each block has at least one character stroke direction, and at least one main character stroke direction is selected according to the density of the stroke directions of the characters; wherein the feature value is the at least one main character stroke direction, the stroke direction The system is divided into 0°, 45°, 90° and 135°.

According to another feature of the present invention, the present invention provides an image processing method performed by an image processing system for recognizing a character of a license plate, the method comprising: receiving an image of the license plate; dividing the image of the license plate into a plurality of character regions, each of the character regions including a plurality of pixels to form a character; each character region is subjected to one-direction feature extraction; and each character region after the direction feature is extracted is divided into a plurality of characters And determining a feature value of each block and combining the feature value of each block to identify the character of the character area.

The simplified conversion system first divides each character region into a plurality of pixel regions, and then compares a pixel point distribution in each pixel region with a plurality of preset patterns, and compares each pixel according to the comparison result. The area is classified, and the classification result replaces the pixel point distribution with the corresponding pattern.

Each of the blocks has at least one character stroke direction, and at least one main character stroke direction is selected according to the density of the stroke directions of the characters; wherein the feature value is the at least one main character stroke direction, the at least The direction of a character stroke is divided into 0°, 45°, 90° and 135°.

11‧‧‧ receiving module

12‧‧‧ character segmentation module

13‧‧‧ Identification Module

S14‧‧‧ steps

S111‧‧‧Steps

S112‧‧‧Steps

7‧‧‧ Block

T ‧‧‧ threshold

19‧‧‧ License Plate

S121‧‧‧Steps

S122‧‧‧Steps

S123‧‧‧Steps

S131‧‧‧Steps

S132‧‧‧step sign

S133‧‧‧Steps

W ‧‧‧Width

H ‧‧‧height

191‧‧‧ License plate characters

Figure 1 is a flow chart of a conventional license plate recognition system.

2(A) is a block diagram of an image processing system for license plate recognition of the present invention.

2(B) is a flow chart of an image processing method for license plate recognition of the present invention.

Figure 3 (A) is a schematic diagram of the horizontal projection of the license plate image of the present invention

Figure 3 (B) is a schematic view of the vertical projection of the license plate image of the present invention

4 is a schematic diagram of the license plate image character segmentation of the present invention.

FIG. 5 is a schematic diagram showing the directional feature extraction of the image processing method for license plate recognition of the present invention.

6 is a schematic diagram of the character direction feature processing of the image processing method for license plate recognition of the present invention.

FIG. 7 is a schematic diagram of the block feature processing of the image processing method for license plate recognition of the present invention.

FIG. 8 is a schematic diagram of a coding table comparison of an image processing method for license plate recognition of the present invention.

2(A) is a block diagram of an image processing system for license plate recognition of the present invention. As shown in FIG. 2(A), an image processing system for license plate recognition is used to identify a character 191 of a license plate 19, and the image processing system includes: a receiving module 11 and a word. The component module 12 and the character recognition module 13 may include a camera to receive an image of the license plate 19, as generally known, the license plate 19 At least one character 191, such as a number, an English letter or another character symbol, is attached, and the receiving module 11, the character segmentation module 12, and the character recognition module 13 can include a specific hardware. The device, the processor, or a computer device executing the image processing program, etc., can process the image of the license plate 19 to recognize the character 191 on the license plate 19.

2(B) is a flowchart of an image processing method for license plate recognition according to the present invention. The image processing method for license plate recognition is performed by the image processing system. Please refer to FIG. 2(A) together, the method firstly The receiving module 11 receives the image of the license plate 19 (step S111). In the embodiment, the license plate 19 has an image of the character AAA-1176. After receiving the image of the license plate 19, the receiving module 11 has a binary value. The image of the license plate 19 is obtained (step S112), so that the binarized image of AAA-1176 can be obtained. Secondly, the character segmentation module 12 is used to horizontally project the binarized image of the license plate 19 (step S121). And vertical projection (step S122), and dividing the binarized image of the license plate 19 into a plurality of character regions (step S123), that is, dividing into "A", "A", "A", "-", "1", "1", "7" and "6" have a total of eight character regions; then, the character recognition module 13 is used to identify the characters in each character region, wherein the character recognition The module 13 first performs a character direction feature processing on each character region (step S131) to obtain "A", "A", "A", - a border character of eight characters such as "1", "1", "1", "7", and "6", and then a block feature is performed for each character region (border character) after the directional feature processing Processing (step S132), to divide each border character into a plurality of blocks, determine a feature value of each block, and combine the feature values to identify the character by comparing the code tables (step S133). Finally, after outputting the result of the comparison of the code table, the license plate is completed. Knowledge (step S14).

As shown in FIG. 3(A) and FIG. 3(B), the character segmentation module 12 performs character segmentation by horizontal projection and vertical projection, and the divided characters are as shown in FIG. 4. Wherein, the calculation of the horizontal projection and the vertical projection is in accordance with the following formulas 1 and 2:

In Equations 1 and 2, p _x ( x ) represents the horizontal projection of the binarized image of the license plate 19, and p _y (y) represents the vertical projection of the binarized image of the license plate 19, f ( x , j ) For the number of pixel grayscale values of the binarized image at the same height H , f ( i , y ) is the number of pixel grayscale values of the binarized image in the same width W , where the height H and the width W are two Value the width and height of the license plate image.

In the identification module 13, as shown in FIG. 5, the character direction feature processing extracts the foregoing directional features for each character region according to the 0° direction, the 45° direction, the 90° direction, and the 135° direction. A pixel, for example, takes a pixel area of 3x3 in a character area as an example, when a pixel (0, 0), (0, 1), (0, 2), (1, 0), (1, 1), ( 1,2), or pixels (1,0), (1,1), (1,2), (2,0), (2,1), (2,2) are image points, then this The center of the pixel area (1,1) is marked as a pixel in the 0° direction and marked with 0; when the pixel is (0,0), (0,1), (0,2), (1,0), (1) , 1), (2, 0), or pixels (0, 2), (1, 1), (1, 2), (2, 0), (2, 1), (2, 2) are images Point, mark the center (1,1) of this pixel area as a 45° image The prime is marked with 1: when the pixel is (0,0), (0,1), (1,0), (1,1), (2,0), (2,1), or pixel (0) , 1), (0, 2), (1, 1), (1, 2), (2, 1), (2, 2) are image points, mark the center (1, 1) of this pixel area a pixel in the direction of 90° and marked with 2; when pixels (0,0), (1,0), (1,1), (2,0), (2,1), (2,2), Or the pixels (0,0), (0,1), (0,2), (1,1), (1,2), (2,2) are image points, then the center of this pixel region ( 1,1) Marked as pixels in the 135° direction and marked with 3. Through this tag processing, "A" and "A" can be obtained from eight character regions such as "A", "A", "A", "-", "1", "1", "7", and "6". The border characters of eight characters such as A", "A", "-", "1", "1", "7", and "6" are as shown in FIG. 6.

The block feature processing divides the area of each border character into a plurality of blocks 7, as shown in FIG. 7, and then calculates the feature values one by one according to the following formula 3:

In Equation 3, B _s is the total number of pixel grayscale values appearing in a block 7, n _gm is the number of pixel grayscale values in a certain direction, and m is the stroke direction (0°, 45°, 90° and 135°), D _m refers to the density in one direction. Therefore, the stroke direction density of each block 7 is calculated one by one according to the stroke direction. Code _m is the direction feature, which is obtained by D _m greater than a threshold T. For example, the direction density of a block 7 is calculated by Equation 3, D _0° =0%, D _45° = 75%, D _90° = 25% D _135° =0%, after the threshold value T is determined. The direction feature is f (0°)=0, f (45°)=1 f (90°)=0, f (135°)=0, and finally the stroke direction of the main character of the block 7 is 45°. After the main character stroke direction of each block 7 is calculated, it is converted into the feature value of the character and represented by the coding of the two bits.

Finally, the eigenvalues of the obtained two-bit code are compared with the pre-established code table 133, as shown in FIG. 8, and the actually represented characters are obtained, and finally all the aligned characters are output. License plate recognition (step S14).

The above-mentioned embodiments are merely examples for convenience of description, and the scope of the claims is intended to be limited to the above embodiments.

11‧‧‧ receiving module

12‧‧‧ character segmentation module

13‧‧‧ Identification Module

19‧‧‧ License Plate

191‧‧‧ License plate characters

Claims (8)

An image processing system for recognizing a character of a license plate, comprising: a receiving module for receiving an image of the license plate; and a character segmentation module for dividing the image of the license plate into a plurality of character regions Each character region includes a plurality of pixels to form a character; a character recognition module is used to identify the character in each character region; wherein the character recognition module first uses each character The meta-region performs one-direction feature processing, and then divides each simplified character region into a plurality of blocks and determines a feature value of each block, and associates the feature values to identify the character; The directional feature processing first divides each character region into a plurality of pixel regions, and then compares a pixel point distribution in each pixel region with a plurality of preset patterns, and compares the results according to each The pixel area is classified, and the classification result replaces the pixel point distribution with the corresponding pattern. The image processing system of claim 1, wherein each block has at least one character stroke direction, and at least one main character stroke direction is selected according to the density of the stroke directions of the characters. The image processing system of claim 2, wherein the feature value is the at least one main character stroke direction. The image processing system of claim 2, wherein the at least one character stroke direction is divided into 0°, 45°, 90°, and 135°. An image processing method is performed by an image processing system to recognize a character of a license plate, comprising: receiving an image of the license plate; Dividing the image of the license plate into a plurality of character regions, each of the character regions including a plurality of pixels to form a character; performing a directional feature processing on each of the character regions; simplifying each of the converted character regions Dividing into a plurality of blocks; determining an eigenvalue of each block, and combining the eigenvalue of each block to identify the character of the character region; and wherein the directional feature processing system first uses each word The meta-area is divided into a plurality of pixel regions, and then a pixel point distribution in each pixel region is compared with a plurality of preset patterns, and each pixel region is classified according to the comparison result, and the pixel result is classified. The distribution is replaced by the corresponding pattern. The image processing method of claim 5, wherein each block has at least one character stroke direction, and at least one main character stroke direction is selected according to the density of the stroke directions of the characters. The image processing method of claim 6, wherein the feature value is the at least one main character stroke direction. The image processing method of claim 6, wherein the at least one character stroke direction is divided into 0°, 45°, 90°, and 135°.