KR101774735B1 - Method for improvement of shadow area around car number plate and car number recognition - Google Patents

Abstract

The present invention is to solve a problem of an existing method. Provided is a method for improving a shadow area around a vehicle number plate and recognizing a vehicle number wherein an image with high quality can be obtained by removing a shadow around a vehicle number plate. According to an embodiment of the present invention, provided is a vehicle number recognition system comprising: at least one camera which is installed on a parking garage and obtains an image including a number plate of a vehicle; and a control unit which recognizes a number of the vehicle from the image obtained by the camera. The control unit corrects the image through a retinex algorithm, recognizes the number of the vehicle, and obtains a final image by generating a retinex uniform layer, a retinex low layer, and a retinex high layer for each mode through the retinex algorithm and multiplying the retinex uniform layer, the retinex low layer, and the retinex high layer.

Description

METHOD FOR IMPROVEMENT OF SHADOW AREA AROUND CAR NUMBER PLATE AND CAR NUMBER RECOGNITION [0002]

The present invention relates to a shadow area enhancement around a license plate and a car number recognition method. And more particularly, to a shadow area enhancement around a license plate and a vehicle number recognition method capable of obtaining high quality images by removing shadows around the license plate.

In the modern society, the economic development and thus the demand for the vehicle as the main transportation means are continuously increasing, and the management personnel and the facilities for maintaining the rapidly increasing population and policing are insufficient. There are a lot of researches and developments in connection with reasons to overcome limited resources and management system, and many intelligent systems using CCTV are being developed. Intelligent illegal parking regulation system and intelligent parking guidance system are being developed and various intelligent systems should be recognized.

Such a car number recognition technology can be applied to various fields such as vehicle control, enforcement, and parking induction, and the higher the recognition rate, the higher the reliability of the intelligent management system.

High-quality images with high recognition rate are essential, and the distribution of low-light areas and high-intensity areas should be evenly distributed. The distribution of the illuminance region is affected by the brightness of the surroundings, or the distribution is different according to the image brightness imbalance according to the weather condition, the camera position, the illumination position, and the vehicle exterior information.

In particular, vehicle license plates can be seen differently depending on constraints such as the position of the camera from the indoor lighting, changes in the external shape of the vehicle, and fixed background objects in the solar environment environment in the outdoor environment. In this case, the brightness of the upper and lower ends of the vehicle license plate is different from that of the image input from the camera.

FIG. 1 is an image taken from a camera for photographing a parking surface and having different illuminance distributions on a license plate. In FIG. 1, when the vehicle license plate is largely divided into two areas, it is divided into a low-illuminance area and a high-illuminance area by the brightness of the surroundings, thereby indicating that the original image is distorted.

The images shown in FIG. 1 may be misrecognized or unrecognized due to differences in brightness during the number recognition process. In the case of (a) and (b), the distribution of the illuminance region changes due to the sunlight, the camera position, and the appearance of the car, which can be improved through image processing.

In the image distortion correction, since the shadow region of the license plate usually has a low-illuminance distribution, in the case of the image taken on the parking surface, it is similar to the color of the parking surface, The information value of the car number may be lost or distorted. This is because the binarization process is highly dependent on the threshold value. For this reason, there is a need for a method of correcting a license plate by generating image distribution images having different illuminance values in one image and blending them in order to obtain high quality images by removing shadows around the license plate. This can be applied to applications where a vehicle identification system such as a parking guidance system is used.

Fig. 2 is an exemplary image in which shadows are generated on license plates in indoor and outdoor environments.

In a parked vehicle, a shadow may appear on the license plate as shown in FIG. 2 due to sunlight, internal illumination, and rear curvature of the vehicle.

This phenomenon occurs mainly due to the structural characteristics of the vehicle and the location of the lighting installed above the vehicle, which is mainly found in the rear part rather than the front part of the vehicle. In the case of a plate with distorted information due to shadows, the number recognition rate is lowered. In order to process distorted information of the image, gamma correction, histogram equalization, retinex algorithm, There are several image processing algorithms.

FIG. 3 is an exemplary image showing a result of image processing through a gamma correction, a histogram smoothing, and a Retinex algorithm on a license plate image having a low low-illuminance distribution, FIG. 4 shows an example image of a license plate image having a low illuminance distribution and a high illuminance distribution, Smoothing, and retinex algorithm.

First, the gamma correction or histogram smoothing algorithm for the global processing of the whole image has the following characteristics. In the case of gamma correction (B), gamma is used to indicate the degree of contrast in the image, which is the contrast correction of the contrast in the image. However, since each value is differently applied to each environment, it is difficult to adaptively adapt to environmental changes. Therefore, as shown in FIG. 3, it can be seen that the low-illuminance area in the license plate area is not improved in the case of the gamma corrected image. In addition, as shown in FIG. 4, it can be seen that the gamma correction improves the low-illuminance distribution of the license plate area slightly.

On the other hand, the histogram smoothing (C) means that the pixel values of the image shifted to a certain portion are uniformly dispersed. However, the histogram smoothing result of the image varies according to the threshold value of each image and the region where the histogram of the original image is biased. Therefore, as shown in FIG. 3, the histogram smoothing also shows that the low-illuminance area is not improved, but the pixel values are smoothed and the contrast is improved. In addition, as shown in FIG. 4, in the case of histogram smoothing, most of the pixel values exist in the low-illuminance region and are smoothed and forcibly unfolded, so that the image is greatly distorted.

That is, since both algorithms show linear characteristics and depend on the system, it is difficult to obtain a desired result image for each input image.

Also, since the Retinex algorithm (D) has a nonlinear characteristic different from the gamma correction algorithm or the histogram smoothing algorithm, the resulting image value varies depending on the input image.

In particular, the Retinex algorithm has the following features. Unlike a camera, it is an algorithm that adapts to the surrounding environment and improves the contrast by adapting the characteristic that the human eye can adapt to the time and distinguish color information about the object even in low light conditions. Therefore, the Retinex algorithm has the advantage of showing better results than the global image processing method.

However, there is a problem in that a high quality image can not be obtained due to the characteristics of the separated image of low light intensity and high light intensity. That is, as shown in FIG. 3, the illuminance area is improved, but the bright area is saturated and the recognition of the letter area is somewhat lowered. In addition, since the resolution of the original image is not high, noise and the like are also emphasized in the image processing result, and it can be confirmed that the image distortion is generated as a whole. Also, as shown in FIG. 4, it can be confirmed that the low-illuminance image in the license plate area is improved, but it is confirmed that noise is formed in the vehicle part by improving the contrast.

SUMMARY OF THE INVENTION Accordingly, it is an object of the present invention to provide an improved vehicle license plate around a vehicle license plate which can acquire a high-quality image by removing shadows around the license plate, And to provide a recognition method.

It is to be understood that both the foregoing general description and the following detailed description of the present invention are exemplary and explanatory and are not intended to limit the invention to the precise form disclosed. It can be understood.

According to an aspect of the present invention, there is provided a vehicle navigation system including a control unit for recognizing a vehicle number from at least one camera that acquires an image including a license plate of a vehicle and an image acquired by the camera, The control unit recognizes the number of the vehicle by correcting the image through a retinex algorithm, and the control unit controls the retinex uniform layer, the retinex low layer, and the retine And a final image is acquired by multiplying a retinex uniform layer, a retinex low layer, and a retinex high layer by generating a nex high layer, and the vehicle number recognition system according to an embodiment of the present invention .

Also, in one embodiment, the controller may multiply the retinex uniform layer, the retinex low layer, and the retinex high layer through Equation 1 to obtain a final image.

[Formula 1]

는 레티넥스 유니폼 레이어,

는 레티넥스 로우 레이어,

는 레티넥스 하이 레이어를 나타낸다.)(here,

The Retinex uniform layer,

Is a retinex low layer,

Represents the retinex high layer.)

Further, in one embodiment, the control unit may set scale, nscales and variant as parameters in the Retinex algorithm.

Also, in one embodiment, the control may have a scale of 200 to 250 in the parameter.

Further, in one embodiment, the control section may have an enqueue value of three in the parameter.

Further, in one embodiment, the control unit may have a value of 1.0 to 1.5 in the parameter.

Also, in one embodiment, the license plate may be a rear license plate.

Further, in one embodiment, the low-illuminance region in the license plate region constituted by the low-illuminance region and the high-illuminance region may be constituted by 40% to 60% of the license plate region.

Further, in one embodiment, the low-illuminance region in the license plate region constituted by the low-illuminance region and the high-illuminance region may be constituted by 10% to 20% of the license plate region.

According to an embodiment of the present invention, it is possible to obtain a high-quality image by removing shadows around the license plate.

It should be understood, however, that the effects obtained by the present invention are not limited to the above-mentioned effects, and other effects not mentioned may be clearly understood by those skilled in the art to which the present invention belongs It will be possible.

BRIEF DESCRIPTION OF THE DRAWINGS The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this application, illustrate preferred embodiments of the invention and, together with the description, And shall not be interpreted.
1 is an image taken from a camera for photographing a parking surface, an example image having different illuminance distributions in a license plate,
FIG. 2 is a diagram showing an example image in which shadows are generated on license plates in indoor and outdoor environments,
FIG. 3 is a diagram showing an example image showing image processing results through gamma correction, histogram smoothing, Retinex algorithm,
FIG. 4 shows an example image showing a result of image processing through a gamma correction or histogram smoothing, a retinex algorithm for a license plate image having a low illuminance distribution and a high illuminance distribution,
FIG. 5 is a diagram showing a Retinex uniform layer image generated by a Retinex algorithm for a license plate image having a low low-
FIG. 6 is a diagram illustrating a retinex low-layer image generated through a Retinex algorithm for a license plate image having a low low-
FIG. 7 is a diagram illustrating a retinex high-layer image generated by a Retinex algorithm for a license plate image having a low low-
FIG. 8 is a diagram illustrating a result of multiplexing the layer images of FIGS. 5 through 7,
9 is a diagram showing a Retinex uniform layer image generated through a Retinex algorithm for a license plate image having a low illuminance distribution and a high illuminance distribution similar to each other,
FIG. 10 shows a retinex low-layer image generated through a Retinex algorithm for a license plate image having a low illuminance distribution and a high illuminance distribution similar to each other,
FIG. 11 is a schematic diagram of a retinex high-layer image generated by a Retinex algorithm for a license plate image having a low illuminance distribution and a high illuminance distribution similar to each other,
FIG. 12 is a diagram showing a result of multiplexing the layer images of FIGS. 9 to 11,
13 is a flowchart illustrating a vehicle recognition method according to an embodiment of the present invention.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings so that those skilled in the art can easily carry out the present invention. However, the description of the present invention is merely an example for structural or functional explanation, and the scope of the present invention should not be construed as being limited by the embodiments described in the text. That is, the embodiments are to be construed as being variously embodied and having various forms, so that the scope of the present invention should be understood to include equivalents capable of realizing technical ideas. Also, the purpose or effect of the present invention should not be construed as limiting the scope of the present invention, since it does not mean that a specific embodiment should include all or only such effect.

The meaning of the terms described in the present invention should be understood as follows.

The terms "first "," second ", and the like are intended to distinguish one element from another, and the scope of the right should not be limited by these terms. For example, the first component may be referred to as a second component, and similarly, the second component may also be referred to as a first component. It is to be understood that when an element is referred to as being "connected" to another element, it may be directly connected to the other element, but there may be other elements in between. On the other hand, when an element is referred to as being "directly connected" to another element, it should be understood that there are no other elements in between. On the other hand, other expressions that describe the relationship between components, such as "between" and "between" or "neighboring to" and "directly adjacent to" should be interpreted as well.

It should be understood that the singular " include "or" have "are to be construed as including a stated feature, number, step, operation, component, It is to be understood that the combination is intended to specify that it does not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, or combinations thereof.

All terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs, unless otherwise defined. Commonly used predefined terms should be interpreted to be consistent with the meanings in the context of the related art and can not be interpreted as having ideal or overly formal meaning unless explicitly defined in the present invention.

1st  Configuration of Embodiment

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings.

에 들어있는 배경성분

을 제거하는 것이다. 입력영상의 평균적인 영상을 배경영상으로 생각할 수 있으며, [수학식 1]과 같이 나타낼 수 있다.First, the basic principle of the Retinex algorithm will be described as follows. The Retinex algorithm uses the input video

Background ingredient in

. The average image of the input image can be regarded as a background image, and can be expressed as [Equation 1].

를 컨벌루션 한 결과이다. 이때 적당한 크기의 필터사이즈를 갖는 가우시안 필터를 적용하여 얻을 수 있다. 필터를 적용하면 입력영상에서 필터 사이즈보다 작은 스케일은 무시하는 효과를 줄 수 있다.Here, the background image includes a Gaussian filter

. At this time, it can be obtained by applying a Gaussian filter having a filter size of an appropriate size. Applying a filter has the effect of ignoring scales smaller than the filter size in the input image.

Here, the Gaussian filter can be defined as the following equation (2).

는 표준 편차로 영상의 블러링 양을 조절한다. 이 값의 크기에 따라 영상의 다이내믹 레인지 압축과 디테일의 차이가 생기게 된다.In Equation (2), A is a normalization coefficient for making the maximum value of the Gaussian function 1,

Adjusts the amount of blurring of the image with a standard deviation. Depending on the magnitude of this value, there is a difference in dynamic range compression and detail of the image.

은 다음의 [수학식 3]과 같이 입력영상을 앞서 구한 배경영상으로 나누면 된다.The reflection component of the input image irrespective of background illumination

Can be obtained by dividing the input image by the previously obtained background image as in the following Equation (3).

은 앞서 구한 반사성분에 로그 값을 취해 다음의 [수학식 4]와 같이 얻을 수 있으며, 이는 로그 특성으로 인해 분포범위(다이내믹레인지)를 압축하는 효과를 얻을 수 있다.In addition, the Retinex output

Can be obtained as the following equation (4) by taking the logarithm of the reflection component obtained previously, and it is possible to obtain the effect of compressing the distribution range (dynamic range) due to the logarithmic characteristic.

Here, the SSR (single-scale retinex) algorithm is applied to the retinex algorithm for one scale, and the algorithm is applied to each of the R, G, and B channels for the color image. The SSR algorithm estimates the illumination component of the input image using a Gaussian function based on the log operation. By removing the illumination component estimated from the input image, the reflection component is obtained, and the resultant image is obtained by applying color restoration, gain and offset.

 In addition, when the Retinex algorithm generates images at various scales and outputs the results as one, it is called a multi-scale retinex (MSR) algorithm. At this time, an appropriate weight is given to each single scale, and a result image can be generated as shown in the following Equation (5).

는 스케일에 대한 가중치를 나타내며,

은 출력영상이다.

은 각각의 스케일이 적용된 영상들을 의미한다.At this time,

Represents the weight for the scale,

Is an output image.

Are images to which respective scales are applied.

The Retinex images thus obtained have a range of pixel values between [0,255] through proper offset and stretching. In this case, the offset is a constant, and the stretching is the pixel value of the image image is between 0 and 255. If the actual minimum value and the maximum value are not 0 and 255, the minimum value is set to 0 and the maximum value is set to 255, Relocate. The Retinex output image of the color image tends to be grayed as a whole, so that it is complemented by the following Equations (6), (7) and (8)

,

,

는 R, G, B 각각의 채널 별 스케일이 적용된 영상을 의미하며, C는 상수이고

,

,

는 입력영상의 RGB 채널을 나타낸다.At this time

,

,

Denotes an image to which a scale for each channel of R, G, and B is applied, C is a constant

,

,

Represents the RGB channel of the input image.

In the present invention, a retinex multiply is a result obtained by changing the mode of the Retinex algorithm to obtain result layer images having different characteristics, and multiplying each of the images. The following equation (9) represents the formula of the algorithm of the final image proposed by the present invention.

,

,

는 각각 레티넥스 유니폼(retinex uniform), 레티넥스 로우(retinex low), 레티넥스 하이(retinex high)로 각 출력 이미지를 나타낸다. 각각의 레이어 이미지들은 레티넥스 모드(retinex mode)를 통해 출력되는 결과로 가우시안 필터의 크기에 따라 각각의 특징을 갖는 출력 결과를 나타낸다 레티넥스 로우 영상은 필터의 크기가 작아 영상 내 특징점이나 에지 정보를 강조하며, 레티넥스 하이 영상은 필터의 크기가 커 영상을 전체적으로 블러 시키지만 색상 정보를 많이 갖는다. 그리고 레티넥스 유니폼 영상은 중간 정도의 결과를 생성한다. 3개의 레티넥스 레이어 영상을 이용하여 멀티플라이(multiply) 하는 이유는 각각의 영상을 단순히 더하는 것과 다르게 각각의 컬러정보와 에지 정보를 증폭시켜 나타내기 위함이다. 각 영상에서 가지고 있는 컬러정보와 에지 정보들이 하나의 영상으로 합쳐 원본 영상보다 개선된 결과를 얻을 수 있게 된다. 각 영상들을 곱한 후 255의 제곱으로 나누어주는 것은 영상에서 표현하는 픽셀 값의 최대치는 255이기 때문에 영상이 멀티플라이 될 때마다 픽셀 값들을 255로 나누어 정규화 해주는 과정이 필요하기 때문이다.

,

,

Represent each output image with retinex uniform, retinex low, and retinex high, respectively. Each layer image is output through a retinex mode. As a result, each layer image shows output characteristics of each feature according to the size of the Gaussian filter. The retinex low image has a small filter size, The retinex high image has a large size of the filter and blurring the image as a whole but has a lot of color information. Retinex uniform images produce moderate results. The reason for multiply using three retinex layer images is to amplify each color information and edge information differently from simply adding each image. The color information and the edge information of each image are combined into one image to obtain an improved result than the original image. Since the maximum value of the pixel value represented by the image is 255, the process of normalizing the pixel values by dividing the pixel values by 255 is necessary every time the image is multiplied.

In addition to the mode, the Retinax algorithm has three parameters, including scale, nscales and variant, as parameters that determine the result of the image.

First, Enscale is a parameter that determines whether Retinex mode is used in image restoration of Retinex algorithm. The minimum value is 1 to 2 and the maximum value is 3. At 3, Retinex mode can be used to acquire one result image from Retinex Uniform, Retinex Low and Retinex High images.

When the minimum value is 1, the scale distribution is determined as one median filter, and the scale value is divided by half. When the value is 2, the scale filter is determined as a median scale and a maximum scale, and a result image is obtained using each scale value.

The scale represents the scale of the retinex algorithm, and the mean value of the significant parameter changes from min. 0 to max. Value 250. As the value decreases, the image tends to be grayed, and as the value increases, the color information tends to saturate. The scale has the best results at 240.

Finally, the variance is a parameter that affects the saturation of the resulting image. The mean value of the significant parameter changes was from 0 to the minimum value of 3.0. The lower the value, the higher the saturation. The higher the saturation, the lower the saturation. The variant gives the best result at 1.2.

FIG. 5 shows a retinex uniform layer image generated by the Retinex algorithm for a license plate image having a low low-illuminance distribution, FIG. 6 shows a retinex low-layer image generated by a Retinex algorithm for license plate images with low low- FIG. 8 shows a resultant image image obtained by multiplexing the layer images of FIGS. 5 to 7, FIG. 9 shows a low-illuminance distribution and a high-illuminance distribution 10 is a retinex low-layer image generated by a Retinex algorithm for a license plate image having a low illuminance distribution and a high illuminance distribution similar to each other, Fig. 11 is a diagram showing a retinex uniform layer image generated by a retinex algorithm, For license plate images with similar distribution and high intensity distribution, The Retinex high layer image generated by an algorithm, Figure 12 is the result of the video image by multiplying the image a layer of 9 to 11.

As described above, first, as shown in FIGS. 5 to 8, an image to be used as a multi-layer is generated as an intermediate image of each of the original images using the Retinex algorithm, which is generated on the basis of the following criteria.

Here, the original image is an image including a license plate of a vehicle obtained through at least one camera provided in the parking lot, wherein the license plate is preferably a rear license plate of the vehicle.

Retinex algorithms are Retinex Uniform, Retinex Low, and Retinex High. Gain filtering is performed according to the size of each scale. Retinex low image emphasizes edge information and feature information in each image, and Retinex high image produces a lot of color information. Also, the uniform method produces an intermediate result between the two methods mentioned above. By using the three layer images generated in this way, meaningful information of each image is combined into one image, and improved results can be obtained through the result image. Here, the license plate images shown in Figs. 5 to 8 may be composed of a low-illuminance area and a high-illuminance area, and the low-illuminance area may be 10% to 20% of the license plate area.

That is, even if the resolution of the image is low, it is possible to obtain more improved results than the result of the single-layer-based image processing because each significant image component is combined into one image.

As shown in FIGS. 9 to 12, in the retinex low result image, a clear license plate image can be confirmed as a result of a clear contrast of the license plate area. In the Retinex high result, the license plate image is somewhat saturated, It can be seen that the image color information is improved. The results of Retinex uniforms can be seen in the middle of these two results. In addition, it is possible to confirm the improved image as shown in FIG. 12 by multiplexing each result image. 9 to 12, the low-illuminance area may be configured to be 40% to 60% of the license plate area in the license plate area composed of the low-illuminance area and the high-illuminance area.

The image quality loss information can be evaluated through the maximum signal-to-noise ratio (PSNR) with respect to the resultant image obtained through the image processing of the original image as in the above-described embodiment.

Here, the maximum signal-to-noise ratio can be defined as the following equation (10).

In this case, MAX is a maximum value of the image, which is obtained by subtracting the minimum value from the maximum value of the corresponding channel. For example, for an 8-bit grayscale image, it is 255 (255-0). MSE stands for mean square error, which means mean square error. The measured value of the difference between pixel values indicates the difference between pixel values of the original image and the output image. However, PSNR can not be defined because MSE is 0 for lossless images.

The following [Table 1] and [Table 2] show the result of comparing video quality using PSNR. Numerical values are shown between the original image and each retinex mode output image and the resultant image of the multiplying result. As a result, we can see that the Retinex multiplication method has a higher PSNR value than the single layer method.

[Table 1] shows the PSNR of the retinex result of the license plate image with low low-illuminance distribution, and Table 2 shows the PSNR of the retinex result of the image with low illuminance distribution and high illuminance distribution similarity.

Original video retinex uniform retinex low retinex high retinex multiply PSNR 9.5dB 9.4dB 9.6dB 16.2dB

Original video retinex uniform retinex low retinex high retinex multiply PSNR 12.3dB 11.8dB 11.7dB 13.9dB

Second  Method of embodiment

13 is a flowchart illustrating a vehicle recognition method according to an embodiment of the present invention.

Referring to FIG. 13, a vehicle recognition method according to an embodiment of the present invention includes an image acquisition step (S10) and an image correction step (S20).

In the image acquiring step (S10), a license plate is photographed through at least one camera (not shown) provided in the indoor or outdoor parking lot, and an image including the license plate is acquired.

In the image correction step S20, as described above, the image obtained through the camera is changed in mode through the Retinex algorithm to generate a resultant image having three characteristics (retinex uniform, retinex low, and retinex high). Herein, the retinex low image emphasizes the edge information in the image according to the size of the Gaussian filter, and the retinex high image emphasizes the color information in the image. And retineex uniform image has edge information and color information evenly.

The controller multiplies the three images to improve the image of the distorted area by the low resolution image or the environmental factor, obtains the high quality image, and performs the brightness correction of the shadow area of the license plate.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The foregoing description of the preferred embodiments of the invention disclosed herein has been presented to enable any person skilled in the art to make and use the present invention. While the present invention has been particularly shown and described with reference to preferred embodiments thereof, it will be understood by those skilled in the art that various changes and modifications may be made therein without departing from the spirit and scope of the invention as defined by the appended claims. For example, those skilled in the art can utilize each of the configurations described in the above-described embodiments in a manner of mutually combining them. Accordingly, the present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

The present invention may be embodied in other specific forms without departing from the spirit or essential characteristics thereof. Accordingly, the above description should not be construed in a limiting sense in all respects and should be considered illustrative. The scope of the present invention should be determined by rational interpretation of the appended claims, and all changes within the scope of equivalents of the present invention are included in the scope of the present invention. The present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein. In addition, claims that do not have an explicit citation in the claims may be combined to form an embodiment or be included in a new claim by amendment after the filing.

Claims (12)

A recognition system for recognizing a number of a vehicle parked in a parking lot,
At least one camera installed in the parking lot for acquiring an image including a license plate of the vehicle; And
A controller for recognizing the number of the vehicle on the image acquired by the camera; Lt; / RTI >
The control unit recognizes the vehicle number by correcting the image through a retinex algorithm,
The controller generates the retinex uniform layer, the retinex low layer, and the retinex high layer for each mode according to the retinex algorithm using the retinex algorithm. The retinex uniform layer, Wherein the final image is obtained by multiplying the retinex low layer and the retinex high layer.
The method according to claim 1,
The control unit
Wherein the final image is obtained by multiplying the retinex uniform layer, the retinex low layer, and the retinex high layer through the following Equation (1).
[Formula 1]

(here,

The Retinex uniform layer,

Is a retinex low layer,

Represents the retinex high layer.)
The method according to claim 1,
The control unit
And sets scale, nscales and variant as parameters in the Retinex algorithm.
The method of claim 3,
The control unit
In the parameter
Wherein said scale has a value between 200 and 250. < RTI ID = 0.0 > 18. < / RTI >
The method of claim 3,
The control unit
In the parameter
Wherein the encecle has a value of three.
The method of claim 3,
The control unit
In the parameter
Wherein the variant has a value between 1.0 and 1.5.
The method according to claim 1,
Wherein the license plate is a rear license plate.
The method according to claim 1,
In a license plate area consisting of a low-illuminance area and a high-
Wherein the low-illuminance area comprises 40% to 60% of the license plate area.
The method according to claim 1,
In a license plate area consisting of a low-illuminance area and a high-
Wherein the low-illuminance region comprises 10% to 20% of the license plate area.
A method for recognizing a car number using a car number recognition system according to claim 1,
A first step of acquiring an image including a license plate of a vehicle parked in the parking lot using the camera; And
A second step of recognizing the vehicle number by correcting the image through a retinex algorithm in the control unit; Lt; / RTI >
In the second step, the controller generates a retinex uniform layer, a retinex low layer, and a retinex high layer for each mode using the retinex algorithm, Wherein the final image is obtained by multiplying the retinex uniform layer, the retinex low layer, and the retinex high layer.
11. The method of claim 10,
In the second step,
Wherein the final image is obtained by multiplying the retinex uniform layer, the retinex low layer, and the retinex high layer through the following Equation (1).
[Formula 1]

(here,

The Retinex uniform layer,

Is a retinex low layer,

Represents the retinex high layer.)
11. The method of claim 10,
In the second step,
Wherein scale, nscales and variant are set as parameters in the Retinex algorithm.

Images