Embedded System Design for Night-time Vehicle Detection Based on Headlight Motion and Tracking

• Main Authors: Hung-Jr Wu, 吳弘智
• Other Authors: Ming-Hwa Shen
• Format: Others
• Language: zh-TW
• Published: 2013
• Online Access: http://ndltd.ncl.edu.tw/handle/56895307869269353466

碩士 === 國立雲林科技大學 === 電子與光電工程研究所碩士班 === 101 === Traffic surveillance is very important with the rapid development of the city. Modern daytime vehicle identification systems have been quite perfect. But nighttime vehicle recognition systems has many difficult problems to overcome, because there are many other sources of illumination in nighttime, including vehicle lights, street lamps, and ground reflect lights. These light sources make it very difficult to identify vehicle objects. This paper presents the reflected light filtering algorithms based on modified histogram equalization to solve these interference of light sources, so that we can recognize the vehicle headlights among light sources. Finally, we will track these vehicle lights to classify the vehicle correctly. Experimental results show that the proposed algorithm can filters the reflected lights very well whether it is sunny or raining at night. The vehicle identification accuracy can achieve at least 90% which is better than the latest works. The execution speed of our vehicle identification is 35 FPS in computer.