系统工程与电子技术
繫統工程與電子技術
계통공정여전자기술
Systems Engineering and Electronics
2015年
12期
2683-2688
,共6页
邱昊%黄高明%左炜%高俊
邱昊%黃高明%左煒%高俊
구호%황고명%좌위%고준
多目标跟踪%机动目标%标签多伯努利%序贯蒙特卡罗
多目標跟蹤%機動目標%標籤多伯努利%序貫矇特卡囉
다목표근종%궤동목표%표첨다백노리%서관몽특잡라
multi-target tracking%maneuvering target%labeled multi-Bernoulli (LMB)%sequential Monte Carlo
针对标准标签多伯努利(labeled multi-Bernoulli,LMB)算法只考虑了单个运动模型的问题,提出了一种适用于跳转马尔科夫系统的多模型标签多伯努利(multiple model LMB,MM-LMB)算法。首先对目标状态进行扩展,将多模型思想引入 LMB 算法得到了新的预测和更新方程,并给出了算法的序贯蒙特卡罗实现。仿真实验表明,MM-LMB 算法能对多机动目标进行有效跟踪,在复杂探测环境下跟踪精度优于多模型概率假设密度(multiple model probability hypothesis density,MM-PHD)算法和多模型势平衡多目标多伯努利(multiple model cardinality balanced multi-target multi-Bernoulli,MM-CBMeMBer)算法;所提算法计算量当目标相距较远时低于MM-PHD 和 MM-CBMeMBer,目标聚集时增长速度快于对比算法。
針對標準標籤多伯努利(labeled multi-Bernoulli,LMB)算法隻攷慮瞭單箇運動模型的問題,提齣瞭一種適用于跳轉馬爾科伕繫統的多模型標籤多伯努利(multiple model LMB,MM-LMB)算法。首先對目標狀態進行擴展,將多模型思想引入 LMB 算法得到瞭新的預測和更新方程,併給齣瞭算法的序貫矇特卡囉實現。倣真實驗錶明,MM-LMB 算法能對多機動目標進行有效跟蹤,在複雜探測環境下跟蹤精度優于多模型概率假設密度(multiple model probability hypothesis density,MM-PHD)算法和多模型勢平衡多目標多伯努利(multiple model cardinality balanced multi-target multi-Bernoulli,MM-CBMeMBer)算法;所提算法計算量噹目標相距較遠時低于MM-PHD 和 MM-CBMeMBer,目標聚集時增長速度快于對比算法。
침대표준표첨다백노리(labeled multi-Bernoulli,LMB)산법지고필료단개운동모형적문제,제출료일충괄용우도전마이과부계통적다모형표첨다백노리(multiple model LMB,MM-LMB)산법。수선대목표상태진행확전,장다모형사상인입 LMB 산법득도료신적예측화경신방정,병급출료산법적서관몽특잡라실현。방진실험표명,MM-LMB 산법능대다궤동목표진행유효근종,재복잡탐측배경하근종정도우우다모형개솔가설밀도(multiple model probability hypothesis density,MM-PHD)산법화다모형세평형다목표다백노리(multiple model cardinality balanced multi-target multi-Bernoulli,MM-CBMeMBer)산법;소제산법계산량당목표상거교원시저우MM-PHD 화 MM-CBMeMBer,목표취집시증장속도쾌우대비산법。
For the problem that the standard labeled multi-Bernoulli (LMB)filter only considers the single motion model case,a multiple model LMB (MM-LMB)filter for maneuvering target tracking is proposed.By introducing the jump Markov (JM)system to the LMB method,the extended recursion formulations are presen-ted,and the sequential Monte Carlo implementation of the proposed method is given.Simulations show that the MM-LMB filter can track multiple maneuvering targets effectively,and has higher tracking accuracy than the multiple model probability hypothesis density (MM-PHD)filter and the multiple model cardinality balanced multi-target multi-Bernoulli (MM-CBMeMBer)filter in complex detection environment.The calculation cost of the proposed method is lower than MM-PHD and MM-CBMeMBer when the targets are not closed,while grows faster than the compared algorithms when the targets gather together.