基于均值平移和自适应预测的运动目标跟踪

针对运动目标跟踪算法不足之处,提出结合改进的均值平移与自适应预测的目标跟踪算法,基于Bhattacharyya系数值进行Kalman滤波器与粒子滤波器之间的切换.引入Kalman滤波器为Mean Shift算法估计初始点,在跟踪稳定的情况下进行模板更新,根据Kalman残差大小判定是否发生遮挡：部分遮挡情况下即结合Kalman滤波器实现对快速运动目标的实时跟踪;完全遮挡情况下结合均值平移和粒子滤波进行鲁棒跟踪.实验证明,改进的算法可以有效地提高跟踪算法的效率,并且能很好地解决遮挡问题.

Moving Target Tracking Based on Modified Mean Shift and Adaptive Prediction

A moving target tracking method with modified Mean Shift and adaptive prediction is pro- posed to improve the performance of the current Mean Shift tracker, which bases on Bhattacharyya value to achieve the switches between Kalman filter and Particles. Kalman filter is introduced to esti- mate initial point for MeanShift tracker, the template in the tracking stable situations is updated. Ac- cording to the kalman poor estimate the procedure of occlusion： when there is part-occlusion, the Mean Shift tracker with Kalman prediction is used to achieve real time performance in fast target tracking; the Mean Shift tracker with particle prediction is used to achieve robust tracking performance. Experi- ments indicate the modified algorithm can effectively improve the tracking efficiency and solve the oc- clusion problem.