基于改进粒子滤波的多目标跟踪算法研究

Research of multi-targets tracking algorithm based on improved particle filtering

针对粒子滤波多目标跟踪中数据关联和估计问题,将经典粒子滤波扩展成在给定几个观测过程时多目标状态过程的估计。用Gibbs采样作为估计和分配关联向量的主要方法。目标状态向量和关联概率被联合估计而没有经过列举,修剪、门限等操作,避免了合并的弊端。测试算法已用于检测目标状态的变化,包括纯方位目标和实际的视频序列,在较为复杂的跟踪条件下,也能实现稳定跟踪。实验结果表明,该算法有较强地解决数据关联问题的能力。

Focus on data association and estimation problems in particle filtering multi-targets tracking,classical particle filtering is extended into multi-targets state estimation in the given several observation process.Gibbs sampling is regarded as the methods of estimation and distribution correlation vector.Target state vector and association probability is jointly estimated without list,trim,threshold and other algorithms.Merger drawbacks is avoided.Test is running in bearings-only target and real video sequence.Stable tracking is realized under the complex tracking conditions.Experiments show that algorithms have strong ability of solving data association problems.