The Design And Implementation Of Automatic Vehicle Location And Identification System

With the development of national economy and traffic, license plate automatic recognition technology of LPR(License Plate Recognition) plays more and more important role in the intelligent transportation. It includes license plate location, character cutting and character recognition technology.This thesis research a variety of license plate recognition technology deeply, the mainly work for this thesis as follows:Introduce the situation for abroad License Plate Recognition research, the latest research status.For preprocessing stage, firstly, converter the color image to gray image, then de-noise for the gray image and use gray stretch and image enhancement techniques to enhanced the useful information from the gray image in order to improve the recognition rate of the character, the last of the processed image binarization, to improve the efficiency of post-character recognition.In order to get the positioning plate region, this thesis uses the properties of multi-resolution analysis of wavelet transform, the horizontal and vertical projection, mathematical morphology segmentation of license plate images.On character segmentation, combined with domestic standard license plates, as well as positioning plate image wavelet transform coefficients of high frequency sub, using its vertical projection of the entire license plate segmentation, extract each character to be identified, and the split normalized character, image recognition processing easy.Finally in the stage of character recognition, using HU seven invariant moments to extract the character feature. Rotation, scaling, translation invariance of HU moments has better robustness, the extracted features to characterize the nature of effective license plate. Error back propagation network is a kind of nonlinear mapping of neural network, the input layer, hidden layer neurons, after the treatment, think the money spread to the output layer and get the results, by modifying the middle layer neuron weights, gradually reduce the error of fitting, and then re forward propagation process. So repeatedly, until the error is less than a predetermined threshold. This thesis uses the HU seven invariant moments as the license plate recognition, BP neural network as the classifier, and the use of MATLAB and neural network design for automatic license plate recognition system of vehicle license plate recognition and has high recognition rate.