摘要
首先研究了Mean Shift算法(均值漂移算法),针对Mean Shift算法在跟踪视频运动物体应用中的不足,提出了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,对运动物体跟踪有了明显的改进,有了更好的鲁棒性。
This paper studied on the Mean Shift algorithm. Due to the problem of Shift Mean algorithm in tracking moving objects,this paper proposed a weighted fusion algorithm through Kalman filtering prediction window and three frame difference method to extract the window to generate a new detection window as a mean shift algorithm for search window. At the same time,the bandwidth of the kernel function and the target model were updated. The experimental results show that the optimized algorithm tracking moving objects is significantly improved and has better robustness.
出处
《信息技术》
2017年第1期127-130,共4页
Information Technology
