一种改进的基于核跟踪的运动物体跟踪算法

摘要

针对核跟踪中均值漂移算法候选目标模型起始检索窗口选择的不足，将卡尔曼滤波预测的窗口和三帧差法提取的窗口通过加权的方式，产生一个新的检测窗口作为均值漂移算法的检索窗口，同时对核函数带宽和目标模型进行更新。实验表明经过优化的算法，对运动物体跟踪的效果有了明显的提升，算法的鲁棒性也得到了增强。%Aiming at the deficiency of the initial retrieval window in the candidate target model of mean-shift algorithm , a new detection window is produced as the retrieval window of mean-shift algorithm by weighting the window of kalman filter and extracted window of three image difference methods .At the same time ,the bandwidth of kernel function and the arget model are updated .Experiments show that the optimization algorithm has been significantly improved after the detection of moving object ,the robustness of the algorithm has been enhanced .