Research On License Plate And Vehicle License Optical Character Recognition

With the development of transportation, the application study of vehicle license plate(VLP) and vehicle license(VL)’s optical character recognition(OCR) has become more important in the domains of computer vision and smart transportation. In this research project, the author designed two programs, OCR of VLP and OCR of VL. With no addition of other devices, these two programs are able to automatically identify the information of VLP and VL by machine rather than manual operation. This indicates the success of automatic inspection and management of vehicle information, and traffic management automation. With the development of information and communications technology, automatic extraction, recognition and output characters from the VLP and VL optical images, becomes one crucial step in the promotion of modern traffic control system automation.This project outlined and summarized the current research status of OCR in China and abroad. The author then described how to design two programs. Three algorithms have been chosen and applied in both programs, which are information extraction(IE), character segmentation(CS), and character recognition(CR).The first part of the project, IE, means to extract character strings out of an optical image of a VL. After pre-possessing of the image, the standardized photo will be cropped. The segments which contain required information of VL will then be further processed by support vector machine(SVM). Only one most appropriate VL segment will be selected. The next part, CS, aims at dividing the obtained VL segment into several standardized small picture segments, which will only contain one individual character. These picture segments are required to be processed by slant correction algorithm, for improving the recognition accuracy and speed.To the last part of the project, with the applications of two algorithms, neural network and pattern matching, the two programs will finally produce the most appropriate character. By comparing the above two algorithms and SVM, the practicality of each algorithm has also been analyzed. With the use of recognition theory, this project has designed two OCR programs of both VL and VLP, and the obtained characters have successfully arrived at the anticipated accuracy rate.