Research And Implementation For License Plate Recognition Technique Of Parking Lots

With the rapid development of economy in our society, there are more and more ofprivate cars in our country. This is a challenge to the management of traffic securityand parking lots, so the realistic problem of traffic management in the future needs tobe solved as soon as possible. The license plate number, as the unique identity ofvehicles in car park, can be automatically registered and validated by the technologyof license plate recognition. Therefore, the technology will reduce the workload ofmanagers greatly and achieve intelligent management of parking lots as the importantimpetus in future. The main content of this paper is to research on the key technologyof license plate recognition for parking lots.In the paper, we have mainly researched the key technology of license platelocation, character segmentation and character recognition, and proposedcorresponding algorithms. Before license plate location, the captured images arepre-processed, including image graying, image enhancement, image filtering, imagebinarization, etc; then a license plate location method based on the number of binaryimage jump and mathematical morphology is proposed. Compared to the existedlocation methods, the performance of the proposed method is quite good for stronganti-interference ability and higher accuracy. After the location of the vehicle image,the extracted license plate region is continued to be processed by adjusting tilt licenseplate, cropping the rim and rivet, and unifying the background of binary image. And acharacter segmentation method based on the connected regions of license platecharacters is used to divide the license plate into seven single character images. At thesame time, the processing method of combination and split is used when thecharacters are adhesive and separate. Lastly, all the single characters are normalized,and some common methods for character recognition are analyzed in detail, thencharacter recognition methods based on template match and character recognitionmethod based on BP neural network are improved, which increased the recognitionaccuracy and velocity. After the recognition of the vehicle image, the result isbroadcasted by the recorded voice and preserved in case of necessary lookup.In this paper, the location accuracy has been increased by the proposed licenseplate location algorithm, and the algorithm of character recognition has beenimproved. In addition, some problems have been solved effectively in character segmentation, the purpose of license plate recognition for parking lots has beenrealized，and all of the algorithms are simulated in Matlab(R2007b).