摘要
设计了一种基于特征点的图像跟踪算法.采用卡尔曼滤波器跟踪运动目标,并针对跟踪过程中的旋转及遮挡问题,提出相应策略:根据特征光流计算得到目标的旋转角度,适时地更新特征点的匹配模板,解决了旋转问题;当有局部遮挡发生时,通过模板与图像的匹配相关系数判断出被遮挡的特征点,并把这部分特征点加以滤出,而只有未遮挡特征点的跟踪结果送到卡尔曼滤波器,从而有效解决了局部遮挡问题.试验结果表明,这种跟踪算法具有跟踪精度高、鲁棒性强的特点.
An image tracking algorithm based on distinguished points is presented. The Kalman filter is used to tracking moving targets. In view of the problems of the rotation and occlusion, some strate gies are proposed： Firstly, the rotation angle of the target can be calculated by optical flows of distinguished points. Then matching masks used in tracking are updated based on rotation angle, and the rotation problem is solved. When partial occlusion occurs, occluded distinguished points will be found out according to the low correlation coefficients between masks and the image, and then given up. Only tracking results from unoccluded distinguished points are sent to the Kalman filter. So the partial occlusion problem is also solved. The experiments show that this tracking algorithm has many advantages, such as high tracking precision and robust behaviour.
出处
《武汉理工大学学报（交通科学与工程版）》
2006年第3期398-400,462,共4页
Journal of Wuhan University of Technology（Transportation Science & Engineering）
基金
湖北省科技攻关项目(批准号2002AA102B01)
武汉理工大学校基金项目资助(批准号:xjj2005176)
关键词
特征点
跟踪
旋转
遮挡
distinguished points
tracking
rotation
occlusion
