Correlation based method of candidate verification for license plate extraction

A novel candidate verification algorithm for PLE is proposed in this paper. In the new method, auto-correlation curve and projection properties are combined to verify the candidates. Considering different Chinese car plate, the framework of proposed candidate verification method is constructed. Firstly, a pre-process which gets the accurate vertical boundaries of the character in the plate is performed. Then a three-step process is used to extract different types of car plates. The output of the proposed method is the binary car plate which has done slant correction and can be sent to the segmentation and recognition stages directly. Experimental results show the efficiency of proposed method.