相机运动条件下的行人检测与跟踪  

Pedestrian detection and tracking from moving cameras

作　　者：李春明[1] 许云峰[1] 尚燕[1] 武少广[1] 

机构地区：[1]河北科技大学信息科学与工程学院,河北石家庄050000

出　　处：《计算机工程与设计》2013年第9期3333-3337,共5页Computer Engineering and Design

基　　金：河北省科技厅基金项目(11213591);河北教育厅基金项目(2010239);河北科技大学校立基金项目(XL201162)

摘　　要：行人检测与跟踪在司机辅助安全系统和视频监控等领域具有重要的地位。针对目前存在的关键问题,如人体运动,相机运动,背景及形状、角度等变化对检测及跟踪带来的干扰,提出了一种将运动信息与形状信息相结合的行人检测方法,准确检测运动摄像机拍摄的直立运动人体;使用了基于小面积目标的跟踪算法进行人体跟踪;利用实际拍摄的视频序列进行算法验证。实验结果表明,混合检测算法速度快,准确率高;基于小面积的跟踪算法能够鲁棒的跟踪检测到的运动人体。Pedestrian detection and tracking are important to many advanced systems such as driver assistance system, visual sur- veillance system and so on. Due to the motion of objects, the camera and the background and to the variations in appearance, pose and background clutter, it is very difficult to achieve an ideal robustness. A detector to detect the standing and moving hu- man in videos with moving cameras is presented, which is based on the combination of motion information and human shape in- formation. Then, the moving person is tracked based on a small area tracking method. The approach is tested on real video ex- periments. From the experimental results we can see that the new detector provides a quick and robust detection. And our sys- tem is an integrated system which can perform person detection as well as tracking.

关 键 词：运动目标检测与跟踪 运动人体检测 HOG MSaM 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]