中文摘要：

为改善人体在跟踪过程中部分遮挡情况下的表现,提出了一种具有鲁棒性的联合跟踪方法.先利用显著图算法从人体区域中获取基于特征的感兴趣区域（ROI）,然后用所提出方法对其进行跟踪.跟踪方法结合了差值平方和准则与基于颜色的均值漂移两种跟踪器的优点,使用卡尔曼滤波器进行状态估计,同时将由显著图算法产生的联合显著点作为局部多区域代替全局区域对人体感兴趣区域进行跟踪.试验结果表明,所提出的跟踪方法在人体部分被遮挡情况下的表现有了明显的改善,适合实时跟踪运动人体,具有很好的鲁棒性.

英文摘要：

In order to improve the performance in tracking moving people when they were partially occluded,a new joint tracker of good robustness was proposed.The regions of interest（ROI） from human motion area were firstly obtained by an improved saliency map algorithm,and then ROI was tracked by the method proposed.The advantages of sum of squared differences（SSD） and color-based Mean-Shift（MS） tracker were adopted in the joint tracker.Kalman filter was used to evaluate the state.Simultaneously,the joint points which were produced by joint saliency map algorithm were used to track ROI instead of the global ones.Results of experiments show that the joint tracker is better when the ROI is partially occluded.The method is suitable for real-time tracking the regions and possesses a good robust performance.