The Research And Implementation Of License Plate Recognition System Under Complex Background

Automatic license plate recognition technology was introduced to China from the end of twentieth Century, and mainly inludes the process of license plate location(LPL), character segmentation(CS) and character recognition(CR) with the help of image processing technology. At present, the domestic license plate recognition(LPR) systems perform well in a single scene. However, due to the influence of environment, illumination change and image capture angle in complex scenes, the vehicle license plate image has low resolution and complex background. These factors make it difficult to locate the license plate, on the other hand, license plate characters’ adhesion or breaking will result in license plate not accurately segmented and identified and finally effect the overall performance of the vehicle license plate recognition system.Through the research analysis, this paper using domestic typical blue bottom white lettering plates and yellow bottom black lettering plates as experimental objects, optimizes the results of license plate location, character segmentation and character recognition under complex background,and improves the performance of license plate recognition system under complex background to a certain extent. Aspects of main work has been done in this paper as follows:(1) after the coarse position of the license plate, a line LP accurately located by character- like connected components filtered by the horizontal center line of all detected connected components, so effect of license plate accurate positioning is improved;(2) before the license plate character segmentation, using support vector machine(SVM) to extract license plate hog and HSV histogram features to determine the license plate color, yellow or blue.The classification learning has been proved effective in inverting the LP color in this system;(3)over the character recognition stage, through the character-centered image processing, using LibSVM+sift to successfully realize the automatic recognition of license plate image. Finally, based on the C++ and Opencv library, this paper designs and implements the whole system, which improves the performance of the vehicle license plate recognition under complex background in a certain extent.