License Plate Recognition Technology Research

Image-based license plate recognition ( LPR ),which can promote traffic management efficiency and make it more scientific,standardized and intelligentized,is an important application of computer vision and pattern recognition in the field of ITS.Generally,LPR system mainly comprises modules of plate location ,character isolate and character recognition.In this thesis,we did a lot of research on LPR,brought forward some effective strategies and completed some experiments.Based on these we developed a LPR software.(1) In the part of plate location and segmentation,we designed and implemented a method through edge detection and plate region search.We discussed edge detec-tion,region search strategy and region confirm strategy in detail.This method turned out to be accurate,fast and coincident with the actual facts.(2) In the part of character isolation,we adopted a horizontal segmentation method by edge detecting and jump-searching,a vertical segment method grounds on projection and transcendental knowledge.We also designed a region-based segmentation arithmetic to refine the isolation result.(3) In the part of character recognition,we simply introduced pre-process method of character image and ordinary character recognition methods at fist,then projected a three level character classifier.They are coarse classifier based on fuzzy logic,pinpoint classifier based on the matrix element comparison idea we proposed and similar character classifier based on special point characteristic.In the dissertation,we also discussed image contrast enhancement,slant rectification and image binary,which are difficulty points in LPR. According to the traits of license plate images,we designed a linear transform function to perform contrast enhancement.Using hough transformation and image reverse rotation, a slant rectification method was proposed. We researched text document image binarization methods normally used and selected methods of K-mean clustering and LEVBB .Beside these,image interpolation methods and their application in image rectification and zoom were also discussed.