基于单目视觉的夜间车辆识别与测距方法分析-analysis of night vehicle identification and ranging method based on monocular vision.docxVIP

优秀毕业论文 精品参考文献资料 Abstract Night driving safety has always been a worldwide problem. To protect drivers and vehicles, most vehicles are equipped with many safety features such as safety belts, airbags, etc, which can reduce the damages and injuries during collision accidents. However, these passive safety techniques cant fundamentally avoid traffic accidents, active safety has become the hotspot, it can predict and avoid dangers, night driving safety early warning technique is a solution. This paper mainly studied the vehicle identifying and ranging method at night based on monocular vision, it is one of the basic studies of the night driving safety early warning, and how to finish the vehicle identifying and ranging work exactly and quickly decides the performance of the early warning system. The major research contents of this essay are as follows: Vehicle identifying method at night. A real-time vehicle identifying method based on taillight images at night in RGB space was given. To extract taillight color feature, (R-G) color characteristic was chose as the input, and adaptive threshold algorithm was used to segment image; the taillights were paired by defining pairing rules and difference degree; the taillights conglutination and vehicle overlapping problems were solved. Vehicle ranging method at night. Firstly, defined the coordinate system; and then chose Zhang Zhengyou plane camera calibration method; finally, built a vehicle ranging model by the pinhole imaging theory and priori knowledge. System implementation. Vehicle identifying and ranging system was realized by the Matlab image processing toolbox. This paper’s characteristics and creative contributions are as following: To the complex traffic at night, (R-G) color characteristic is used as the input of night vehicle image segmentation, taillight areas is successfully extracted. The taillights conglutination and vehicle overlapping problems is solved. A vehicle ranging model is built based on pinhole ima