【作者】 王芝； 徐晓滨； 文成林；

【Author】 Zhi Wang,Xiao-bin Xu,Cheng-lin Wen College of Automation,Hangzhou Dianzi University,Hangzhou 310018

【机构】 杭州电子科技大学信息与控制研究所；

【摘要】 本文在随机集的框架下给出了多目标跟踪的滤波模型,介绍了由此推广而来的PHD滤波器,并在线性混合高斯情况下给出了它的解析表达式。在实际使用过程中,PHD方法存在两个问题,一是计算量随目标个数和递归步数呈指数级增加,二是无法得到目标的运动轨迹,也不能确定目标的"身份"。本文针对不同的问题,给出相应的解决办法:优化算法可以大大降低计算量,"聚类"分析可以确定目标的运动轨迹及"身份"。仿真结果也说明了本文方法的有效性。更多还原

【Abstract】 In this paper,a filter model about Multi-target Tracking is present under the random set framework.Then the PHD filter is used to process the model and its closed-form is given under the linear Gaussian mixture situation.There exist two problems in PHD method.First is that the calculation is very heavy and increases exponentially.Second is the method can not identify the target and its trajectory.In order to solve the problems above,an optimized algorithm is shown to release the heavy load of calculating in PHD and a cluster analysis method is given to identify the target and its trajectory.In the last of the paper,the simulation is used to prove the efficiency of the method.更多还原