Automotive Grade Based On The Characteristic Information Of The Video Image To Extract Identification

Along with the development of computer,the computer vision and pattern recognition become an important research topic.The major purpose of using computer to process the video image is to understand and recognize the image automatically.Then through the dynamic image sequence processing,the computer can get more useful information.License Plate Recognition(LPR)is a special computer vision system focusing on the plate recognition and has become one of the most popular research topics.So this paper is to do the research on the image feature extraction from video and do experiment on license plate recognition to verify the effectivity.The main research contents are as follows:(1)Detect the moving vehicle by background subtraction,and present a background set-up method through modified multi-frame mean.Also reduce the computational complexity of background subtraction by setting small detect region.(2)The accurate of license plate location affects directly the following parts of the recognition. On the basis of some plate location algorithm using plate feature,the thesis adopts a method based on the mathematical morphology and geometric features of plate to coarsely locate the plate,and based on the gray level jump of the texture feature of plate image to accurate position.(3)The thesis presents a plate trace method by using matching algorithm based on the feature of edge and plate texture,in which we also combine the least square method to estimate the movement path of the plate.(4)Use a modified project method to segment the characters.Also implement the character recognition by neural network,which uses 3-level BP network and is based on the 13-feature extraction method.In a word,this paper designs a whole license plate recognition system.Firstly it detects moving car from video images.Then it locates the plate in the key frame and uses the plate template to match and track the license plate in the following images.At last,choose the legible plate to recognize the characters.And we get satisfied test result of our plate recognition system in Matlab simulation.