摘要
针对单步延迟无序量测条件下多目标跟踪中,概率假设密度滤波对目标数量与状态估计误差偏大问题,提出了后向预测高斯混合概率假设密度滤波算法(BP-GMPHD)。该算法在后向预测框架内,以高斯混合概率假设密度滤波器为基础滤波算法,计算各高斯分量的回溯状态并进行再更新,经剪枝与合并等步骤获得最终的目标数量与状态估计。仿真验证表明,该算法在无序量测条件下保持了良好的滤波性能,能够准确估计多目标数目和状态。
In multi-target tracking process with one-step-lag out-of-sequence measurement (OOSM) , probability hypothesis density may cause low estimation precision for target number and state. To solve the problem, a backward prediction Gaussian mixture probability hypothesis density (BP-GMPHD) filtering was proposed. Within the backward prediction framework, taking Gaussian mixture probability hypothesis density as basis fil-tering algorithm, backtrack state of each Gaussian component was calculated to obtain target number and state estimation after pruning and merging, etc. Simulation results showed that the proposed algorithm could effectively keep a good filtering performance with OOSM and accurately estimated the multi-target number and state.
出处
《探测与控制学报》
CSCD
北大核心
2017年第3期118-123,共6页
Journal of Detection & Control
关键词
多目标跟踪
无序量测
单步延迟
高斯混合
后向预测
multi-target tracking
OOSM
one-step-lag
Gaussian mixture
backward prediction