基于图像特征点匹配的车辆运动检测  

Vehicle Moving Detection Based on Image Feature Point Matching

作　　者：武林[1] 彭复员[1] 赵坤[1] 

机构地区：[1]华中科技大学电子与信息工程系

出　　处：《仪器仪表学报》2005年第z1期581-582,共2页Chinese Journal of Scientific Instrument

基　　金：国家自然基金(60475024);国家863计划(2002AA401001-4B)资助项目

摘　　要：提出了应用于交通视频监控系统中车辆运动检测和分割的一种新颖有效的方法。采用了较为简易的特征点匹配检测车辆并利用纹理识别对车辆进行分割。此方法不需要对序列图像进行预处理,能有效消除光照阴影和背景,实验结果表明该方法可以实时准确提取道路交通信息。An efficient method is presented for vehicle moving detection and segmentation in traffic video monitoring system. Moving vehicles in a scene are detected by using a feature point matching method, and vehicles in a motion window are segmented by using a texture segmentation method. This method does not call for preprocessing of video sequence images, yet shadows in lighted conditions and background information can be safely and effectively discarded. Experimental results demonstrated that the method is robust in that traffic information can be extracted in real-time fashion.

关 键 词：运动检测 特征点匹配 纹理分割 视频监控系统 

分 类 号：TH7-55[机械工程—仪器科学与技术]