Research On License Plate Character Recognition Algorithm

In this era of rapid development, private car seems to be essential goods, which leads to the rapid growth in the number of motor vehicles, but also brought a lot of problems to the traffic management. In order to solve these problems as soon as possible, the research of vehicle intelligent management system is also in constant development and improvement. The role of the license plate as our identity card,which is a sign of difference between the identity of other vehicles, in order to design a better intelligent traffic management system, license plate recognition becomes a focus of the study. So the value of recognition techniques used in the system is very large.This paper mainly studies on the algorithm used in license plate recognition system, which including license plate location, character segmentation, and character recognition algorithm. For the algorithm used in locating, due to operators of the structure can not change with the license plate size within the image, which may led to the case of failure localization, so using the adaptive morphological localization algorithm. For the process of character segmentation, Chinese characters may not correct segmentation, so using the segmentation algorithm based on projection and characteristics of the delimiter; For the process of recognition, BP network algorithm is likely falling into the "local minimum", so proposed using BP neural network algorithm that improved by genetic algorithm. After a large number of simulation analysis,the proposed algorithm can improve the recognition rate in the premise of reducing the recognition time.First, this paper makes a series of image preprocessing to the acquisition vehicle.In the process of license plate location, combination adaptive algorithm and traditional morphological localization algorithm; it can get the connected region of license plate in the image. Then, the license plate was located successfully which according to some texture features of license plate. Second, this paper studies thelicense plate character segmentation algorithm, it using the segmentation algorithm based on projection and characteristics of the delimiter, which makes license plate segment successful. Later, in order to extract the character feature more accurately, so need to do normalization to the segmented character. For the problem of recognition process of BP neural network algorithm may fall into “local minimum”, and convergence time became long, it use the BP neural network algorithm which improved by genetic algorithm for character recognition. Last, this paper do the experiment simulation and comparative analysis to the improved algorithm used in license plate recognition process, it shows the new algorithm in license plate recognition process is feasible, and is helpful to improve the recognition rate.