Research On Automatic License Plate Recognition Algorithm In Road Video Surveillance

With the development of economic globalization, the number of vehicles is increasing, followed by the traffic jam, exhaust pollution, noise pollution, traffic accidents, etc. The birth of the intelligent transportation system has brought the hope to the full range of high efficiency management of modern traffic. The license plate recognition system is one of the important branches of the intelligent transportation system, which plays an irreplaceable role in practical applications.License plate recognition system refers to a technology of the extraction of vehicles driving on the road in real-time from the monitoring system and automatic identification of vehicle licence information. Generally, license plate recognition system is divided into three parts. The first part is the license plate location, which is to locate the license plate from the image. The second part is the character segmentation, which is to extract each character area from the license plate area.The third part is the character recognition, which is to classify each character correctly. Most of the traditional license plate recognition is for a certain image. The identification accuracy and reliability are not high. This paper emphasizes the importance of vehicle tracking in the vehicle license plate recognition, and makes up for the insufficient of the reliability of the identification of the system.The main work of this paper includes: 1.In the license plate location stage, the improved edge detection method is used to make the initial positioning, which effectively solves the problem of noise sensitive for the edge detection. Then the Ada Boost method is adopted to carry out precise positioning. It is a significant reduction of the heavy workload of the Haar-like features computing tasks. The combination of these methods can improve the accuracy and real-time performance of the location. 2.In the stage of character segmentation, the segmentation error is reduced by using thees methods of combining the connected region and the vertical projection. The two methods are complementary to each other. So that the effect of each character is better. 3.In the character recognition stage, in order to strengthen the recognition of similar characters on structure, this paper proposes a method of using two level classifier. In the first stage, a template matching method based on Euler number is used, and the method of local feature recognition is used in thesecond stage. This method improves the traditional template matching method,which is more efficient and has better recognition effect on the structure of similar characters. 4.A simple voting mechanism is proposed. The identification results of the same vehicle are carried out by using a simple "minority majority" voting method, and the final identification results are given.The methods proposed by each stage are finally tested and verified by experiments. The experimental results show that the algorithm based on vehicle tracking can effectively improve the accuracy of the license plate recognition, and the reliability of the accuracy is stronger.