Adaptive Local Binarization Method for Recognition of Vehicle License Plates

Abstract

A vehicle license-plate recognition system is commonly composed of three essential parts: detecting license-plate region in the acquired images, extracting individual characters, and recognizing the extracted characters. But in the process, the problems like damage of license-plate and unequal light effect make it difficult to detect accurate vehicle license-plate region and to extract letters in that region. In this paper, to extract characters accurately in the license- plate region, a local adaptive binarization method which is robust under non-uniform lighting environment is proposed. To get better binary images, region- based threshold correction based on a prior knowledge of character arrangement in the license-plate is applied. With the proposed binarization method, 96% of 650 sample vehicle license-plates images are correctly recognized. Compared to existing local threshold selection methods, about 5% of improvement in recognition rate is obtained with the same recognition module based on LVQ.