Adaptive Template-matching Method for Recognition of Low-resolution License Number

Abstract

In this paper, we propose an adaptive template-matching method to recognize of low-resolution license number. The first step is to extract license plate from low-resolution image by manual. The license plate is divided into several blocks, and every block corresponds to a license character. Subsequently, we estimate the adaptive filter via pesudoinverse operation, and it is applied to template to generate filtered template which is similar to divided character. Eventually, the filtered template and the similarity measurement between the license character and the template are obtained. According to the similarity measurement, the proposed method can identify the license number. Moreover, our method outputs several possible resultants of license number recognition, it is helpful to police in image forensic. The advantage of our approach is that training process is unnecessary. The quantized rank histogram is measured to evaluate the recognition performance. The experiment results demonstrate that the proposed method is capable of coping with license number recognition in low-resolution images.