一种基于多样性优化的视频目标跟踪方法

摘要

针对粒子滤波算法在视频目标跟踪过程中出现的粒子退化问题,提出一种新的基于多样性优化的粒子滤波跟踪算法.融入基于莱维飞行机制的布谷鸟搜索优化算法,扩大了搜索范围,充分保留了有效粒子的数量和粒子的多样性.在MATLAB 7.0运行环境下进行仿真实验,将该算法与原始粒子滤波算法及基于粒子群优化的粒子滤波改进算法进行状态估计实验对比以及运动目标跟踪实验对比.实验结果表明:该算法具有良好的运行效果,改善了粒子退化与匮乏现象,对视频运动目标可获得较理想的跟踪精度及良好实时性和鲁棒性.%Aiming at the particle degradation problem of particle filter algorithm happening in the process of video target tracking, a new particle filter tracking algorithm is proposed based on diversity optimisation, integrated into cuckoo search optimisation algorithm based on levy flight mechanism.The algorithm expands the search scope as well as fully retains the amount of effective particles and the diversity of particles.Under the environment of MATLAB 7.0 operation, the simulation experiment is conducted, comparing the proposed algorithm with original particle filter algorithm and the improved particle filtering algorithm based on particle swarm optimisation by state estimation experiment and moving target tracking experiment.The results of the experiment show that the proposed algorithm operates well, improving the phenomenon of particle degradation and deficiency, and obtains high tracking accuracy, good real-time performance and robustness for video moving targets.