Study On Algorithms For License Plate Recognition System Under Complex Background

With the economic development, the number of urban vehicles is getting largerand larger. As a result, the traffic problem is becoming worse and worse, whichmakes the development of intelligent transportation system become an urgent problemto solve. License plate recognition system which is an important part of the intelligenttransportation system, has been widely used in peccancy vehicles audit, citymonitoring system, parking lot, highway toll and other occasions related to theauthentication license plate. However, the installations of city vehicle platerecognition system mostly are fixed system. They are installed at every majorintersection whose advantages include simple and stable background, slightdisturbance and high recognition rate. However, it lacks flexibility.Mobile vehicle for traffic surveillance can stay at any place to detect the licenseplate of passing vehicle and get the relevant information through the database, so thevehicle license plate recognition system need to work under complex background. Inview of this problem, the author makes an intensive study of the vehicle platerecognition algorithm under complex background in this paper.First of all, the algorithm uses codebook method to establish background modelfor the video image and detects the position of moving vehicle according to the model.Then, the license plate is positioned fast and accurately by using the mathematicalmorphological method combined with plate texture characteristics and geometricfeatures, and characters are segmented by using projection method according to thegeometrical characteristics of the license plate character. Finally, to avoid thedisadvantages of large operation load and long operation time of the conventionalcharacter recognition method, which cannot meet the real-time requirements, templatematching of feather classification algorithm is adopted to recognize characters on thelicense plates.The hardware developing platform is composed of an Analog camera, DaHengimage acquisition card and PC. VC++6.0and OpenCV visual function library formthe software environment. The experimental results show that the system can locatethe license plate and recognize the characters under complete background correctlywith high stability and high discrimination score. The research demonstratestheoretical significance as well as practical value.