Neural Network Algorithm And Its Application In License Plate Recognition

China is undergoing rapid economic development, the rapid increase in carownership period, as China to develop the infrastructure-transport pathway in roadtraffic is becoming increasingly important. Currently, the topic of traffic managementare to improve transport efficiency, to build safe and stable traffic control system byobtaining electronic information technology management efficiency. In order to enhancetraffic management modernization and intelligent, automatic license plate recognitiontechnology is the key. This technology is better than traditional technology inmanagement efficiency and level, also save manpower and material resources. Sovehicle management is more scientific and standard and make the traffic security.In the field of pattern recognition, neural network in pattern recognition research ispopular. Compared with the traditional pattern recognition, neural network is highlyparallel, non-linear and highly redundant, highly nonlinear, self-study andorganizational skills. As people intelligent traffic control system for license platerecognition accuracy requirements continue to increase, the use of neural networktheory makes sense in license plate recognition.Image segmentation is the key steps and the basic premise of image recognition andimage understanding, image segmentation quality direct impact on the effect of thesubsequent image processing, image segmentation is crucial.This dissertation elaborate the status of the technology of the license platerecognition at home and abroad, on this basis, a series of theory and technology by BPneural networks and image processing technology are applied to license platerecognition, including license plate image graying, license plate image binarization,license plate character segmentation, normalized character image, license plate characterimage feature extraction and so on. BP neural network algorithm optimized are used torecognition characters. And the gradient descent method with momentum training makethe learning process to reduce oscillations and speed up the convergence, BP networkmay recognize the license plate character better.Experimental results showed that the application of this method based on imageprocessing, combined with BP neural network technology, license plate characterrecognition rate reached98.5%.