摘要
针对粒子滤波跟踪算法在视频跟踪中存在的计算复杂、计算量庞大,无法满足实时系统的应用需求,提出了实时粒子滤波跟踪算法。利用粒子滤波器潜在的数据并发特征,在集群环境下,设计并实现了分布式并行粒子滤波跟踪算法,给出了主从模式下的算法设计、数据划分、负载平衡及通信策略。实验结果表明,随着粒子数增加,计算量以幂指数增大,并行跟踪算法的执行时间明显减少,有效地提高了跟踪精度、降低计算时间,能够满足硬实时系统的时间约束。
Real-time application of particle filter tracking algorithm is limited due to their inherent complex and intensive computation.A real-time algorithm based on data parallelization was proposed to achieve on-line tracking using particle filter based on video sequence.The task division,data decomposition and dynamic load balance and communication strategy under Master-Slave model about distributed parallel algorithm were proposed.The simulation result shows that computing time of parallel algorithm is great de...
出处
《系统仿真学报》
CAS
CSCD
北大核心
2009年第18期5651-5655,共5页
Journal of System Simulation
基金
国家自然科学基金(60572061)