Research of Indonesian license plates recognition on moving vehicles

Abstract

The recognition of the characters in the license plate has been widely studied, but research to recognize the character of the license plate on a moving car is still rarely studied. License plate recognition on a moving car has several difficulties, for example capturing still images on moving images with non-blurred results. In addition, there are also several problems such as environmental disturbances (low lighting levels and heavy rain). In this study, a novel framework for recognizing license plate numbers is proposed that can overcome these problems. The proposed method in this study: detects moving vehicles, judges the existence of moving vehicles, captures moving vehicle images, deblurring images, locates license plates, extracts vertical edges, removes unnecessary edge lines, segments license plate locations, Indonesian license plate cutting character segmenting, character recognition. Experiments were carried out under several conditions: suitable conditions, poor lighting conditions (dawn, evening, and night), and unfavourable weather conditions (heavy rain, moderate rain, and light rain). In the experiment to test the success of the license plate number recognition, it was seen that the proposed method succeeded in recognizing 98.1 % of the total images tested. In unfavorable conditions such as poor lighting or when there are many disturbances such as rain, there is a decrease in the success rate of license plate recognition. Still, the proposed method's experimental results were higher than the method without deblurring by 1.7 %. There is still unsuccessful in recognizing license plates from the whole experiment due to a lot of noise. The noise can occur due to unfavourable environmental conditions such as heavy rain.