The Research For License Plate Recognition Technology Based On Classified BP Neural Network

The research and realization on License Plate recognition technology plays an importantrole in the modern traffic.The purpose is to solve the increasingly serious trafficmanagement problems.The LPR system involves numerous discipline domains,such asPattern Recognition and Artificial Intelligence,Computer Vision,Digital ImageProcessing,Artifieial Neural Network etc. The paper is mainly study license platelocalization,license plate character segmentation and license plate character recognition,andusing MATLAB completes the simulation of the recognition of license plate ,and gives theexperimental results.License plate localization has handled the vehicle image with grayscale,again carried onthe median filtering, binary processing, and then used Prewitt operator on edge detection,effectively inhibit the noise influence, then using the algorithm of mathematical morphologyof vehicle license plate location, finally has achieved satisfactory results.License plate character segmentation segmented the seven characters of the locatedvehicle images. In this part , first the located vehicle images completed the tilt correction byusing Radon transformation, and then go to frame operation, finally completed license platecharacter segmentation based on the cluster analysis algorithm .License plate character recognition is the final step of license plate recognitiontechnology, The paper researches neural network-based character recognition technology, itbrings a classified BP neural network -based character recognition method, the Paper hasdesigned three kinds of different classification device, accomplishing license plate characterrecognition.In this part,first normalized the charaeters after character segmentation, thenaccording to their own characteristics of Chinese characters and alphanumeric, Chinesecharacters used rough network and stroke complexity index characteristics of fusion method , Alphanumeric adopted one pixel feature extraction method.Finally,this paper designed BPnetwork of characters, BP network of letters, BP network of digitals and letters ,has achievedvery good recognition effect.