Research Of Related Technologies On License Plate Recognition System

The development of the intelligent vehicle management has attracted more attention in such circumstances that traditional artificial vehicle management system are unable to meet the demand of the rapid increase in the amount of vehicles, parking area and highway construction in China. License Plate Recognition(LPR)has become one of the most important research projects of the communications departments, and has shown its promising application prospect in highway charge, violation detection (electronical policeman) and intelligent parking area etc..A typical LPR system consists of three subsystems including plate location, character segmentation and character recognition, while the different stages of the LPR system are all linked together and the former stage has influnce on the latter.The design and implementation of algorithm in the main stages of the LPR including plate location, character segmentation and character recognition have been researched in this paper.The research of plate location. The image processing is firstly completed,which includes color reduction from color image to greyscale, gray stretch, binaryzation and edge detection. As for edge detection, comparation and analysis has been done to the conventional algorithms, and the improved operator is adopted. And then, a simple and practical plate extraction approach based on the combination of horizontal first-order difference gray level projection and prior knowledge has been proposed. Finally, the Hough transformation is applied to correct the plates'small angle deflect, and good performance has been achieved.The research of algorithm on character segmentation. Firstly, horizontal and vertical border have been removed using vertical projection on the basis of the grey level skipping times, and it has realized the Denoise appling morphology. Moreover, a plate location combined with character width and fixed spacing between characters has been discussed, which has been validated by the experiments.The research of algorithm on character recognition based on BP neural networks. Based on the comparison of the conventional methods for character recognition, the BP neural network is chosen. Targetting the features of the plate character image recognition, the selection of the network structure, improvement and the set of initial conditions have been done. Considering the features that the characters can be separated into three classes including Chinese characters, letters and numbers, together with the regulation of the character distribution, three different character categorizers have been designed, which includes Chinese characters categorizer, letters categorizer, and numbers&letters categorizer. By effective training of the network, a considerable ratio of character recognition has been obtained. Based on above mentioned theoretical research, the C++ programming is applied to realize the algorithms of plate location, character segmentation and character recognition, and the LPR system has been designed. The data comparations and analysis of the test results have shown good performances of the approach proposed in this paper.