An Arithmetic Research Based On Neural Network For Car License Plate Recognition

With the rapid development of domestic traffic management, there are a lot ofsystems, including traffic signal automation management system, intelligence trafficmonitor system, GPS, automation parking toll system, and intelligence traffic system,which are apply to the practice based on the license plate recognition. Becauselicense plate is a important part of vehicle management system, therefore, it isnecessary to research license plate system. The core of arithmetic of license platesystem is how to improve the recognition, and this needs to adapt the environment,light, location the speed of cars, and the requirement of rapid processing.In this paper, my work is focused on the image processing by the emulator. I analyze threeproblems, which are license locating, segmentation and character recognition. Before licenselocating, I adopt artificial neural network (ANN) to preprocess the image, then I get theinformation of license plate in evidence. To locate the license plate, I use the inherent features oflicense plate, and then use mathematical morphology to locate the license plate accurately;According to the inside features of locating license plate, I pop the plate-area out by project thelicense plate; The last step is characters recognition, ANN is the key tool in recognition. So Idesign an improved neural network to segment the characters. I get the whole license platerecognition through these steps. The result of emulator proves that my arithmetic can improvethe rate of recognition, effectively.My work has three innovations. Firstly, I use an improved and automatic PCNNneural network to preprocess the image, and I get a binary image, which has anabundant details, intact edges, then I get lots of candidate areas of license plate,based on the inside features of license plate and a new mathematics morphology.Secondly, when I preprocess the aberrance image, I use an arithmetic based on theline features of characters, then segment the image of characters based on improvedproject features. Thirdly, I improve on BP neural network to improve the rate ofrecognition.The result of emulator indicates that through the innovation and amelioration, my workcan increase the speed of processing, improve the rate of recognition, and get the perfect results.