Research For Wavelet And Neural Network And Its Application In License Plate Recognition System

With computer related technology has been increasingly applied in traffic domain, Vehicle License Plate Recognition system become one important part of Intelligent Transportation System. VLPR system has very obvious economic value and practical significance and has been widely applied to various occasions, like vehicle violation capturing, traffic flow monitoring and parking lot intellectualized management, etc.In this paper, we mainly studied five part of VLPR system, including image preprocessing, license plate location, license plate character segmentation, character feature extraction and license plate character recognition.First of all, denoising and enhancing process is conducted for license plate by combining binary wavelet and mathematical morphology, and specific location of license plate is precise positioning.Then, license plate character segmentation is conducted with vertical projection and normalization processing for segmentation results, and the concept of invariant moment is expounded and the invariant of wavelet moment is in-depth researched and demonstrated, and character feature extraction is conducted respectively by using Hu combined moment and wavelet moment.Finally, wavelet neural network is used for classifier of this system, and its algorithm and programming step are described. MATLAB R2015 a is used as an auxiliary tool for simulation experiments, character features of these two invariant moment are input of classifier and its output shows that the recognition capability of wavelet moment is better than Hu combined moment.