基于粒子滤波架构的实时目标跟踪算法研究  

Research of real-time target tracking based on particle filter framework

作　　者：孟军英[1,2] 刘教民[1] 韩明[1] 王娟[1] 

机构地区：[1]燕山大学信息科学与工程学院,河北秦皇岛066004 [2]石家庄学院计算机系,河北石家庄050035

出　　处：《河北工业大学学报》2013年第4期18-22,共5页Journal of Hebei University of Technology

基　　金：河北省自然科学基金(F2012208004);河北省科技支撑计划(12210807);河北省高等学校科学技术研究项目(ZD20132013)

摘　　要：针对粒子滤波算法运算量大,实时性较差的问题,提出了一种基于粒子滤波架构的实时性目标跟踪算法.通过选取优化的重要性概率密度函数,使粒子分布更接近实际的后验分布.并将Camshift算法融入粒子滤波中,使粒子向真实分布进一步聚类,从而可以用较小的计算代价,保证跟踪的精度.最后,将所设计的算法在交通视频监控中进行实际应用测试,实现了对目标车辆的有效跟踪.The main disadvantage of traditional particle filter algorithm needs massive computation, which leads to poor real-time capability. We propose a real- time- target- tracking algorithm based on particle filter framework combined with EKF and Camshift algorithm. By choosing optimized importance density function from EKF algorithm, the new algorithm can make the distribution of particles more suitable to the actual posterior distribution. Then use the Camshift algorithm to rearrange the particles to let them further cluster to the maximal posterior probability density area of the target state. Experimental results show that compared with the traditional particle filter algorithm or Camshift algorithm, the proposed method can efficiently track the target car and the computational complexity reduce.

关 键 词：目标跟踪 粒子滤波 均值漂移 扩展卡尔曼滤波 智能交通 

分 类 号：TN713[电子电信—电路与系统]