A Study Of Algorithms For License Plate Character Segmentation And Recognition

With the rapid growth of modern economy and society , the number of vehicles has been increasing greatly in recent years in China. But Chinese road traffic environment is not increased correspondingly, and traffic management is also lagging behind, which makes that the traffic pressure is increased in modern cities. In order to more effectively carry out a comprehensive urban traffic management, intelligent transportation system has become the main direction of development of a modern urban road traffic management. The license plate automatic recognition with pattern recognition and computer vision is an important technology in the field of intelligent transportation system and also one of the cores in intelligent transportation system. Its purpose is no need to retrofit vehicles and other special devices. Without changing the state of the movement of vehicles it can achieve real-time automatic recognition of license plates. It provide effective assistance to the automatic management of transportation systems, so the license plate recognition system is one of the most important technologies to realize intelligent traffic management .In generally, license plate recognition system can be divided into three components, which are automatic vehicle location, license plate character segmentation and character recognition. Its research mainly related to the pattern recognition, artificial intelligence, computer vision, digital image processing and many other subject areas. Firstly the key technologies of the license plate recognition are researched in this thesis. Then the character segmentation and recognition algorithms are analyzed. Finally various algorithms are run by mathematical software MATLAB7.0. Experimental results show these algorithms are availability and effectiveness.In this thesis, the main research contents are as follows:①In the license plate character segmentation stage, Otsu method based on virus evolutionary genetic algorithm is proposed to select the threshold of license plates pictures. It can ensure the accuracy of selecting the threshold. Comparative experiments show that the method proposed in this thesis is better than Otsu algorithm in selecting the threshold. Finally, the histogram and prior knowledge of license plates are used to design segmentation algorithm to obtain good segmentation results. This method can reduce errors in character segmentation.②In the license plate character recognition stage, feature extraction using moment invariant and the output value of the network using membership are adopted. Then neural networks based on algebraic algorithm are used to identify characters on the license plate. The shortcomings of the complicated structure of neural networks are avoided and at the same time the advantages of neural networks are used, so the networks have uncertain information-processing capabilities. Experiments confirmed that the method of identification is better. Comparative experiments show neural networks based on algebraic algorithm are better than traditional BP neural networks in character recognition.