Design And Implementation Of Correction Algorithm In License Plate Recognition System Tilt

Vehicle license plate recognition system is the core of intelligent traffic, mainly for major road traffic junction, monitor and guide the traffic.Especially in the large parking lot of the city is playing its role cannot be ignored, mainly responsible for vehicle access control, billing and automatic release.Study finds tilt correction technology in recognition system for number plate recognition rate plays a decisive role.In order to strengthen the management of vehicles and improve the speed of vehicular traffic, adapt to the market demand, for outdoor use is developed for vehicle license plate recognition system brings a higher performance requirements, particularly for slant correction algorithm for real-time performance and compatibility as well as correcting errors require higher.This lesson focuses on slant correction algorithm conducted in-depth research, presents a new feature extraction based on color information and corner fast correction algorithm and applications based on the software implementation of the algorithm.Based on the research of after tilt correction of the classic algorithm in vehicle license plate recognition system, implements three common slant correction algorithm, proposed a new method based on color information and corner feature extraction algorithm for fast correction.Realizing the first common algorithm for projective rotation correction algorithm the algorithm calculation method of main research angles of the through different perspective projection of the variance is calculated to determine the inclination;Two common algorithms implemented for the rotation based on interpolation algorithm of rotation of the original and adds common interpolation algorithms include nearest neighbor interpolation, bilinear interpolation algorithm compensate for pixel information lost in the rotation;Will common of tilt correction algorithm embedded to subject group of license plate recognition system in the, and for actual test, through on test results of compared analysis, eventually proposed a while has positioning and tilt correction function of new algorithm, new algorithm first through color space conversion fast positioning and separation license plate regional, again on the regional for mathematics morphology processing, extraction border and eventually positioning out angle points of location, calculation out angle points of minimum inertia moments, inertia moments minimum of angle that for license plate dip;Paper based on color information and slant correction algorithm for feature extraction of corner, which compared with a projected rotation of rotation and interpolation method processing speed, high recognition accuracy and robustness.Software testing shows that the tilt be limited by location and identification algorithm in the original system, to achieve the desired effect, in realization of correction at the same time, suggestions to improve the positioning algorithm, improved positioning algorithm, using the secondary filter, effectively avoiding the effects of noise on positioning speed and accuracy, the improved positioning accuracy up to 98.9%.Finally the character template is optimized, making identification cannot be processed, or a low recognition rate of the original license plate number to accurately identify and ultimately more than 1%-improved recognition rate.Based on this design slant correction algorithm, in case of good weather to identify low resolution images, recognition rate is over 99.5%, algorithms and software for performance and stability, this software is now used in Dalian in the parking lot of a large company, software is stable, in the field of vehicle license plate recognition is of considerable practical value and promotion prospects.