Research And Implementation Of Algorithms For Vehicle License Plate Positioning & Characters Segmentation Based On Image Analysis

A vehicle license plate (VLP) automatic recognition system is an important part of theintelligent security system of a housing estate, and it is based on digital image processing,pattern recognition and computer vision. Using the unique plate register code, a VLPrecognition system implements automatic recognition of images by computer with videocamera, so as to improve the reliability of security and reduce the labor intensity ofsecurity guards. It also has a good prospect in civil applications.For a vehicle license plate recognition system, the key techniques are the algorithm ofvehicle license plate positioning, the algorithm of character segmentation, the algorithm ofcharacter recognition and images preprocessing. The algorithm of vehicle license platepositioning and character segmentation is the fundamental for character recognition. Thispaper focused on the research and implementation of the algorithms for vehicle licenseplate positioning, character segmentation and related image preprocessing.Based on detailed understanding the demand of application, deeply analyzing somedomestic and foreign products and thoroughly studying and evaluating the existing relatedresearches, this paper has proposed simplified algorithms, and implementing them.Firstly, we studied some image preprocessing algorithms relating to the operationsdunng the prophase of VLP recognition, including the conversion from RGB color space toHIS space, the histogram statistics, the gray level transformation, the image sharpeningfiltering and the smoothing of noise filtering. The paper has given out appropriateimplementation methods.Based on image analysis, we studied the core algorithms for VLP positioning,including the gradient-based algorithm for detecting plate candidates, the Sobeloperator-based algorithm for detecting VLP boundaries so as to determine a correct platearea. In order to improve the efficiency and accuracy, we also discussed other positioning methods based on color image, horizontal projection and vertical projection,Next, we discussed the algorithms of Hough transformation, Slope adjustment, and theMarr-Hildreth edge detector, implementing these algorithms for character segmentation.We proposed a contour based algorithm to detect the height and width of plate characters.By that time, we got a list of characters for recognition.Finally, we discussed some shortages in our research, pointing out the developingdirection in the future.This project is sponsored by Suzhou Vocational University.