Vehicle license plate recognition system palys important applications in traffic intelligent monitoring and management, it is also an applied research hotspot of pattern recognition at home and abroad. This article mainly uses digital image processing, mathematical morphology and neural network technology to prosess license plate image, and makes analysis and research aiming at license plate location, tilt correction, character segmentation, feature extraction and character recognition during the prosess of plate recognition, design a complete vehicle plate recognition system, and simulate them in matlab.First, we study the vehicle coloured license plate location method based on the background and the pixel considering the difficulties of vehicle license image location in different backgrounds and illumination conditions, realized car plate location accurately according to the background subtraction and pixel, and analyzed positioning methods based on geometry analysis after morphological processing, made simulated contrast based on positioning model of coloring book and gray scale image.Second, we discuss template matching character segmentation method based on maximum variance considering serious noise and positioning errors problem, realized the exact license plate character segmentation according to maximum of character pixel variance and template matching principle, then introduced character segmentation besed on anti-color-based binary character projection, simulated the process of plate border elimination and segmentation under the projection pattern, and compare these two segmentation methods.Last, study additional momentum factor and dynamic adaptive learning rate BP classifier improvement method considering that standard BP neural network has slow convergence problem and is falling into local minimum, design improved BP neural network by extracting character feature using feature extraction method and method based on outline and feature fusion of Zernike and Gabor transformation, realized the precise identification of characters, letters, numbers, and calculate the color contribution in the recognition.Experimental results show that the combination of pattern recognition and variety of image processing technology can improve the ability of system identification, and it can meet the requirements of real-time license plate recognition in complex environment, it has theoretical and practical significance. |