Design And Implementation Of Vehicle License Plate Recognition System Based On Spiking Neural Networks

The license plate recognition system is a complex system which involves the technologies of image processing, computer vision, artificial intelligence and pattern recognition and so on. It mainly includes three modules those are the license plate location, character segmentation and character recognition. This paper mainly combines the image processing and the pattern recognition technology and does research on license plate location, character segmentation and character recognition deeply. It simulates part of the algorithms by experiments. Meanwhile on this basis, this paper designs the license plate recognition system which can be implanted in ARM development board and the board is special to intelligent parking management system.This paper presents the license plate location algorithm based on spiking neural networks. The algorithm is mainly composed of two parts namely coarse location and accurate location. Firstly, based on the color feature of license plate, it can achieve coarse location using the spiking neural network; Then using another SNN method to extract the edge of the candidate area. The SNN method which extracts the edge can get a better license plate edge map. So this algorithm has high localization rate, which can locate license plate with various background and is conducive to the subsequent character segmentation and recognition.The character segmentation module is mainly based on the vertical projection and the inherent feature of character. Considering the interval between the second character and the third character is maximum, we take this as the starting point. Then according to the wave crest and trough character projection curve of alternating characteristics, five characters on the right of the starting point can be segmented firstly and two characters on the left of the starting point can be segmented subsequently.The character recognition is based on BP neural network. It uses an improved coarse grid feature and coarse periphery to extract the Chinese characters; It utilizes improved coarse grid feature to extract the characters and numbers. This is conducive to the subsequent training of the BP neural network and also improve the rate of recognition.