Research On The Method Of Vehicle License Plate Recognition In Car Wash

Car washing business has been a rapid development for the growing number of private cars.Ordinary wash inside manually on the vehicle license registration record.Automatic recognition of the license plate can automatically identify the purpose of membership.Special environment,illumination conditions,water disturbance,car wash workers covered on the plate,the image area of the license plate area is small or the license plate slant degree is large.Traditional license plate recognition algorithm can not be applied for the reasons above.Based on the basis of existing research on license plate recognition technology,A improvement program to achieve accurate recognition of license plate for car wash special environment is proposed.Through the analysis in the car wash environment license plate automatic recognition problems,this thesis presents a set of suitable for car wash environment for license plate recognition method.License plate location and strategy of character segmentation are based on multi feature.And the character recognition is based on SVM.In the license plate location,the existence of interference environment is complex.Car wash car wash industry problems prevents to solve this problem.The first step is to the location of the ROI region extraction to reduce interference of background information.To solve the problem of license plate of large tilt,the affine transform of image positioning after correction for image monitoring is needed.To solve the problem of the video license plate area is too small,this thesis proposes a coarse positioning based on connected domain screening and colored point accurate positioning of the search method;Secondly,in the license plate segmentation,according to the collected car vehicle image illumination conditions,water mist interference,low resolution,gray stretch the effect of image enhancement is used in this thesis.To solve the problem of license plate image character projection area can not lead to adhesion separation.A projection method is used in this thesis,this thesis improves the image segmentation by using 3 different templates.In character recognition,due to the structure of the Chinese character itself complex,and car wash internal video shooting angle,leading to some fuzzy distortion in the edge of the license plate,in order to solve this problem,a second recognition method after SVM character recognition results is used.This experiment for 128 the vehicle images in the car wash inside,accurate positioning of the image of a total of 115 can achieved,the accuracy rate was89.10%.For accurate positioning of the license plate,there are 98 complete segmentation of characters,the accuracy rate was 85.21%.Finally,the character recognition method proposed in this thesis can complete the correct recognition of 89 license plate images,the recognition accuracy rate was 90.81%.The problems of license plate recognition in car wash environment can be solved which through the method proposed by this thesis.The analysis of the results of the experimental results and compared with other algorithms,the algorithm with targeting is proofed,with the practical value in the car wash.