基于特征点一致性约束的实时目标跟踪算法

A real-time target tracking algorithm based on the consistency constraint of feature points

基于光流法建立特征点的目标模型和随机采样模型,实现平稳的运动目标跟踪.通过考虑特征点在速度和方向上的一致性约束,学习目标的稳定特征,保持目标模型的稳定性,从而获得稳定的跟踪轨迹.对存在光照变化,目标形变,部分遮挡,高速运动,尺度缩放、旋转,图片噪音、模糊等因素影响下的视频进行仿真对比.结果表明,该算法面对局部非刚体目标和变速运动目标均能达到很好的鲁棒性和实时性.

A novel Lucase-Kanade optical flow based method is proposed for the smooth tracking of moving ob-jects. The proposal includes a target model and a random sampling model. By considering the consistency con-straints in speed and direction and studying object stability ,the proposed method establishes stable feature points for objects. Then, a stable tracking trajectory is achieved. Simulation experiences in comparison with traditional methods were conducted under the situations of illumination changes, target deformation, partial occlusion, high-speed motion, zoom scale, rotation, image noise, and fuzziness. The results show that the proposed method has bet-ter performance of robustness and real-time on many objects which include non-rigid and varying velocity running objects.