KR20160083226A - Apparatus and method for recognizing license plate - Google Patents

Abstract

The present invention relates to an apparatus and a method for recognizing a license plate. The apparatus comprises: a camera which photographs a license plate and generates image information; an image reading part which calculates a photographing angle from a license plate frame of the image information, and reads characters by comparing a unit image per location of the license plate frame and sample image data of a pre-stored corresponding photographing angle; and a database stored by matching the sample image data per photographing angle and corresponding character text information. According to the present invention, a license plate recognition rate can be improved in response to every photographing angle, and thus the apparatus can be easily applied in various environments which are in need of license plate recognition.

Description

[0001] APPARATUS AND METHOD FOR RECOGNIZING LICENSE PLATE [0002]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an apparatus and method for recognizing license plates, and more particularly, to a license plate recognition apparatus and method for easily recognizing license plates photographed with an oblique line.

Cameras, CCTVs, etc. installed on the roads are installed on the upper side or the side surface so as not to interfere with the progress of the vehicle. When the license plate is photographed at such a position, character distortion occurs corresponding to the photographing angle.

In order to recognize the license plate photographed by the oblique line, various character recognition techniques have been proposed. However, when the photographing angle deviates from a certain range, the character recognition rate sharply decreases.

In order to solve this problem, a noise is removed by using a Gaussian filter, a boundary is extracted by applying Laplacian to the result, an area of the license plate is estimated, A technique has been proposed in which each character is recognized and each character is recognized through a neural network character recognizer. However, this technique also has a problem that the recognition rate of the license plate significantly deteriorates due to image distortion when the photographing angle becomes large.

Therefore, there is a need for a method to improve the recognition rate of the license plate even when the shooting angle increases.

Korean Registered Patent No. 10-1182173 (Published on September 12, 2012)

SUMMARY OF THE INVENTION It is therefore an object of the present invention to provide a method and apparatus for extracting a photographing angle from a photographed image and comparing it with sample image data for each photographing angle to easily recognize a license plate And a method for recognizing a license plate of the license plate.

According to another aspect of the present invention, there is provided a license plate recognition apparatus comprising: a camera for photographing a license plate to generate image information; An image reading unit for calculating a photographing angle from the license plate frame of the image information, comparing the unit image for each position of the license plate frame with the sample image data of the photographing angle stored in advance, and reading the characters; And a database for matching and storing the sample image data for each photographing angle and corresponding character text information.

It is preferable that the sample image data for each photographing angle is provided for every 1 degree within a range of 1 to 89 degrees by combining the photographing angle of the camera with respect to the ground and the photographing angle of the camera with respect to the traveling direction of the vehicle .

Wherein the image reading unit comprises: a preprocessing unit for performing filtering including noise removal from the image information; A license plate frame extracting unit for extracting the license plate frame from the preprocessed image information; A photographing angle calculation unit for calculating the photographing angle from the extracted license plate frame; A unit image generating unit for generating the position-specific unit image in the extracted license plate frame; And an image comparing unit for comparing the unit image for each position with the sample image data for the corresponding photographing angle to output a matching character.

Meanwhile, the vehicle license plate recognition method of the present invention includes the steps of: performing filtering including noise removal on image information in a preprocessor; Extracting a license plate frame from the image information in the license plate frame extracting unit; Calculating a photographing angle from the extracted license plate frame in the photographing angle calculation unit; Generating a unit image for each position in the unit image generating unit; And the image comparing unit compares the unit image for each position with the sample image data of the corresponding photographing angle stored in advance in the database and outputs the vehicle number combining the matched characters.

As described above, according to the apparatus and method for recognizing license plate of the present invention, it is possible to improve the license plate recognition rate corresponding to all the photographing angles, so that it can be easily applied to various environments requiring license plate recognition.

1 is a configuration diagram of a license plate recognition apparatus according to an embodiment of the present invention.
2 is a configuration diagram of an image reading unit according to an embodiment of the present invention.
3 is a flowchart of a license plate recognition method according to an embodiment of the present invention.
4 is a view showing a photographing angle of the camera with respect to the ground.
5 is a view showing the photographing angle of the camera with respect to the traveling direction of the vehicle.
6 is a view showing a license plate frame at an angle of photographing of the camera with respect to the traveling direction of the vehicle.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS A vehicle license plate recognition apparatus and method according to the present invention will now be described in detail with reference to the accompanying drawings.

1 is a configuration diagram of a license plate recognition apparatus according to an embodiment of the present invention.

1, a vehicle license plate recognition apparatus according to the present invention includes a camera 1 for photographing a license plate and generating image information, a control unit 2 for calculating a photographing angle from the license plate frame of the image information, A database (DB) 3 for storing the sample image data for each photographing angle and the character text information corresponding to the photographing angle and storing the same, ).

Here, in the database 3, sample image data for each photographing angle corresponding to the angle, letter, number, and character position is stored according to the type of the license plate, and sample image data for each photographing angle is captured by the camera 1 And the photographing angles of the camera 1 with respect to the traveling direction of the vehicle are set in units of 1 degree within the range of 1 to 89 degrees, respectively. Further, the database 3 stores characteristic information of sample image data for each photographing angle.

2 is a configuration diagram of an image reading unit according to an embodiment of the present invention.

Referring to FIG. 2, the image reading unit 2 of the present invention includes a preprocessing unit 21 for performing filtering such as noise removal from image information, a license plate frame extraction for extracting license plate frames from the preprocessed image information, A photographing angle calculation section 23 for calculating a photographing angle from the extracted license plate frame, a unit image generating section 24 for generating a unit image for each position in the extracted license plate frame, And an image comparing unit 25 for comparing the unit image and the sample image data of the corresponding photographing angle stored in advance in the database 3 and outputting the matching character.

3 is a flowchart of a license plate recognition method according to an embodiment of the present invention.

Referring to FIG. 3, when the camera 1 photographs license plates on the front or rear surface of the vehicle to generate image information, the preprocessing unit 21 performs filtering such as noise removal (S1).

When the preprocessing is completed, the license plate frame extracting unit 22 extracts the license plate frame from the entire image of the vehicle (S2).

Subsequently, the photographing angle calculation section 23 calculates the photographing angle from the extracted license plate frame (S3). At this time, the photographing angle extracts the photographing angle of the camera 1 with respect to the ground and the photographing angle of the camera 1 with respect to the traveling direction of the vehicle.

On the other hand, the license plate frame has a predetermined standard, and the prescribed letter / number position is determined in accordance with the standard. In addition, the character / number for which the position is determined is limited to the character set and the font.

Using the frame specification and the limited character set and the position of the font, the unit image generating unit 24 generates a unit image for each position (S4).

When a unit image for each position is generated, the image comparing unit 25 compares the unit image for each position with the sample image data of the corresponding photographing angle stored in advance in the database 3 (S5).

Subsequently, the image comparing unit 25 combines the characters matched as the comparison result and outputs the vehicle number (S6).

[Example]

In the present invention, the stereotyped information of the photographed vehicle license plate, that is, the license plate has the shape of a rectangular frame, is composed mostly of numerals, and utilizes information that a limited character set, limited font, and character / An angle is extracted from the photographed image, and the vehicle license plate is easily recognized through comparison with the sample image data for each photographing angle.

In the present invention, the process of extracting the license plate image from the image information photographed by the camera 1, the process of removing the noise from the extracted license plate image or adjusting the size of the extracted license plate image, , Extracting the photographing angle of the camera 1 with respect to the ground from the extracted license plate image and the photographing angle of the camera 1 with respect to the traveling direction of the vehicle, extracting a unit image per position corresponding to the position of the license plate And comparing the unit image of each position with the sample image data of the corresponding photographing angle to extract a character / number.

On the other hand, in the database 3, the sample image data of the character / number used in the car license plate is taken as a combination of the photographing angle of the camera 1 with respect to the ground and the photographing angle of the camera 1 with respect to the traveling direction of the vehicle Is prepared in advance and compared with the unit image of the license plate photographed by the camera (1).

Figs. 4 to 6 are views showing an example in which an angle between the vehicle and the camera is formed. Fig.

4 to 6, since most of the cameras 1 are installed at a considerable height in order to obtain a large angle of view, the license plates to be obtained in the present invention form the photographing angle of the camera 1 with respect to the ground. This is exemplified by α. In addition, the photographing angle of the camera 1 with respect to the traveling direction of the vehicle has various angles. This is illustrated by β.

The angle? Can be obtained from the distance d between the vehicle and the camera 1 and the height h of the camera 1 as follows. The range of angles obtained from (Equation 1) is 0 ° to 90 °.

? = arctan (h / d) --- (1)

Since the vehicle is not photographed only on the front surface of the camera 1, the vehicle and the camera 1 form an angle beta as shown in Fig. Thus, the license plate has a rhombic shape as shown in Fig. In this case, the angle formed with the horizontal line is β, and the angle rotated clockwise in the horizontal line is denoted by -β and the angle rotated in the counterclockwise direction is changed in the range of 0 ° to 90 °.

On the other hand, for faster and more accurate character recognition, binarization, image enhancement, skewed character correction, and the like are performed before the character recognition.

Binarization is the process of converting a color or black-and-white image into a black-and-white image that represents only 0's and 1's. In the histogram of the brightness distribution of pixels, there is a large peak in the foreground and background, and the intermediate value between the two peaks is used as a threshold value to perform the binarization as shown in (Equation 2).

B (x, y) = 0 if G (x, y) <threshold

B (x, y) = 1 if G (x, y)? Threshold ---- (2)

G: camera input pixel, B: binary-converted pixel

Image enhancement is a technique used to emphasize or suppress selected features to efficiently process the image. Image enhancement techniques include a spatial domain method using a mask and a frequency domain method using a fourier transformation or a histogram modification.

The Laplacian filtering operator like (Equation 3) is used to improve the pattern of the character or character region in the low-resolution brightness image, and also has the effect of eliminating the high frequency misfit.

^{(∇ 2 f) (x,} y) = [f (x + 1, y) + f (x-1, y) + f (x, y + 1) + f (x, y-1)] - 4f (x, y) - (3)

The photographed vehicle license plate is assumed to be vertical, but it can be inclined for various reasons. Thus, the inclination of the characters occurs. Even if the degree of the inclination of the characters is small, the performance of the character recognition system is drastically deteriorated. Especially, the projection profile analysis method and the run length smoothing method based on the fact that the characters are basically composed of the rectangular region among the character structure analysis methods can not be applied to the slanted characters . Therefore, correcting the skew of a character must be performed before the character structure analysis process.

The most commonly used method for measuring the degree of skewing of characters is by using a Hough transform. The Hough transform is basically the same as the method (x, y) which corresponds to the curve of the parameter space (ρ, θ) by finding a line segment at an arbitrary angle and a tilted angle of the line segment Conversion.

? = x cos? = y sin? - (4)

In such a conversion according to (ρ, θ) requires the arrangement (T, R) for accumulating the number of the intersecting generally T = -90 ~ +90, R = (( width of character) ² + (length of the character ) ² ), and θ, which has the largest ρ value in the accumulated array, corresponds to the line segment in the original image.

The structure analysis of a character extracts a geometrical structure of a character from a character image using a prior knowledge in a character having a specific format such as a general character such as a newspaper, a paper, a magazine, or a receipt or a check, After dividing into sets, the characteristics of each block are examined to separate characters into areas to be separated.

In case of character structure analysis for character recognition of license plate, it is not necessary to use complicated technique because it uses simplicity of several license plates used in Korea. However, there are some cases where foreign license plate is used. shall. Character structure analysis methods include link element analysis method, adjacent line density analysis method, run length smoothing method, and projection outline analysis method.

The process of comparing the preprocessed and filtered image data in the image comparing unit 25 with the sample image data according to the photographing angle will now be described.

The letters contain various types of information. Therefore, if the distribution of the information in the collected image is known, character recognition in the image is possible. Information is a knowledge of the state in which an object can be distinguished from multiple objects, reducing uncertainty about the state, and entropy is a measure of the degree of disorder, originally being studied from a different viewpoint than thermodynamic entropy Mathematical expressions are coincident and closely related.

In the present invention, a method of determining information contained in an image is used by using information amount and entropy described in Shannon's information theory. That is, entropy for numbers and letters is stored in advance in the database 3 according to the combination of the photographing angle alpha of the camera 1 with respect to the ground and the photographing angle beta of the camera 1 with respect to the traveling direction of the vehicle , Calculates the entropy of the numbers and letters extracted from the images, and recognizes the letters by comparing them.

The method of calculating the entropy is as follows.

When there is a probability distribution of n events as follows,

Pi> = 0 (i = 1, 2, ..., n), ΣPi = 1

The information amount I and the entropy H are as follows

I = -Σ log2Pi

H = -? Pi log? Pi

The amount of information and entropy for the entire character image are as follows.

I = ΣΣ I [i, j] (i, j is pixel information)

H = ΣΣ H [i, j] (i, j is pixel information)

At this time, if the matching between the image data and the sample image data by the photographing angle is made, the corresponding text text information stored in the database 3 is output as a result.

If no matching is made between the image data and the sample image data by the photographing angle, the process of extracting and comparing the sample image data by the photographing angle with respect to the neighboring angle is repeated.

While the present invention has been particularly shown and described with reference to exemplary embodiments thereof, it is to be understood that the invention is not limited to the disclosed exemplary embodiments, but, on the contrary, is intended to cover various modifications and equivalent arrangements included within the spirit and scope of the invention.

1: camera 2: image reading unit
21: preprocessing unit 22: license plate frame extracting unit
23: photographing angle calculating section 24: unit image generating section
25: image comparing part 3: database

Claims (4)

A camera for photographing a license plate to generate image information;
An image reading unit for calculating a photographing angle from the license plate frame of the image information, comparing the unit image for each position of the license plate frame with the sample image data of the photographing angle stored in advance, and reading the characters; And
And a database for matching and storing the sample image data for each photographing angle and corresponding character text information.
The method according to claim 1,
The sample image data for each photographing angle is obtained by combining the photographing angle of the camera with respect to the ground and the photographing angle of the camera with respect to the traveling direction of the vehicle, .
The method according to claim 1,
The image reading unit reads,
A pre-processing unit for performing filtering including noise removal from the image information;
A license plate frame extracting unit for extracting the license plate frame from the preprocessed image information;
A photographing angle calculation unit for calculating the photographing angle from the extracted license plate frame;
A unit image generating unit for generating the position-specific unit image in the extracted license plate frame; And
And an image comparator for comparing the unit image for each position with sample image data for a corresponding photographing angle to output a matching character.
Performing filtering including noise removal on the image information in the preprocessing unit;
Extracting a license plate frame from the image information in the license plate frame extracting unit;
Calculating a photographing angle from the extracted license plate frame in the photographing angle calculation unit;
Generating a unit image for each position in the unit image generating unit; And
Comparing the unit image for each position with sample image data of a corresponding photographing angle stored in advance in the database, and outputting the vehicle number combining the matched characters.

Images