The Study On The Vehicle License Plate Location And Correction

Automatic license plate recognition technology(ALPR)is a technology that uses optical character recognition to read vehicle license.It is an important part of intelligent transportation system.In the image-based license plate recognition algorithm,including four steps: license plate location,license plate correction,character segmentation and character recognition.This article is mainly on the license plate location and the vehicle license correction.The detailed work contents are as following:In the license plate location technology,this article proposed a set of highly effective and accurate license plate location process.Firstly,the original image is preprocessed by Gaussian filtering and histogram equalization.Secondly,the image gradient feature is extracted by Sobel operator.Thirdly,the candidate region of the license plate gradient is obtained by filtering based on the contours,and then the image is extracted.In addition,convert the original image from the RGB color space to the HSI color space,and then use the threshold segmentation to get a binary image.After removing noise by image morphology,the license plate color feature candidate region is obtained.Finally,the color and gradient candidate region are merged to obtain the final license plate location result.Experimental results show that the proposed method has high accuracy and robustness in complex image background.As for the license plate distortion correction technology,this paper proposed a license plate correction method based on the combination of the rectangle of license plate characters and M-estimator.Firstly,the histogram equalization and threshold segmentation of the license plate location are pre-processed.After that,the circumscribed rectangle and the minimum circumscribed rectangle of the license plate characters are extracted,and noise region can be eliminated.Then,the border of the license plate region and the coordinates of the four vertexes of the license plate region are obtained.Finally,the license plate is corrected by the perspective transformation.The method has low computational complexity and is robust to complex situations such as missing edge information and character sticking.