摘要
针对密集杂波条件下的目标检测与跟踪问题,开展极大似然-概率数据关联(maximum likelihoodprobabilistic data association,ML-PDA)算法优化与实时计算问题研究。在算法层面,通过在极大化对数似然比(log likelihood ratio,LLR)过程中引入粒子群优化(particle swarm optimization,PSO)方法,并进一步提出基于观测引导的PSO播撒粒子方式,提升算法的计算效率;在实现层面,提出基于图形处理器(graphic processing unit,GPU)的PSO实现策略。仿真实验结果说明了基于观测引导PSO算法搜索的有效性。在GPU平台上实现该算法获得显著的加速比,验证了所提出方法具有工程实时性。
The target detection and tracking problems when involved in high dense clutter are addressed. Specifically, we propose to solve the optimization and computation problems of maximum likelihood-probabilistic data association （ML-PDA）. The particle swarm optimization （PSO） algorithm to maximize the log likelihood ratio （LLR） is adopted. We propose to initialize the particles of PSO based on measurements, which improves the computation efficiency. Furthermore, we propose a scheme which allows implementing PSO in parallel on graphic processing unit （GPU）. The efficiency of the proposed algorithm and the parallelized scheme are illus- trated based on simulations.
出处
《系统工程与电子技术》
EI
CSCD
北大核心
2015年第12期2677-2682,共6页
Systems Engineering and Electronics
基金
中国博士后科学基金(2015M572463)资助课题
关键词
检测前跟踪
极大似然-概率数据关联
粒子群优化
并行处理
track before detect
maximum likelihood-probabilistic data association （ML-PDA）
particle swarm optimization （PSO）
parallel processing
