计算高效的分布式多传感器PHD融合方法

Computationally Efficient Distributed Multi-sensor PHD Fusion Method

基于广义协方差交集(GCI)融合理论,提出一种计算高效的分布式多传感器多目标跟踪算法,其中概率假设密度(PHD)滤波器在每个传感器节点运行,进行滤波处理。GCI用于融合多个PHD时,融合密度包括大量融合假设,这些假设随着高斯分量的数量增加呈指数增长。因此,GCI融合在实际运行中往往难以计算。为了提高多传感器融合的运算效率,文中通过距离度量将高斯分量聚类,然后进行孤立。距离度量可计算出目标融合后的密度权重,丢弃权重可忽略不计的融合假设,就能够构建简化的近似密度函数。分析表明,所提出的融合算法相较于传统的GCI融合算法,计算效率能够呈倍数提升。在先后出现12个目标的仿真场景中,通过实验验证了所提融合算法的有效性。

Based on generalized covariance intersection(GCI)fusion theory,a computationally efficient distributed multi-sensor multi-target tracking algorithm is proposed in this paper,in which probability hypothesis density(PHD)filters are run at each sensor node for filtering.When GCI is used to fuse multiple PHDs,the fusion density is consisted of a large number of fusion hypotheses which increase exponentially with the number of Gaussian components.Therefore,in practice GCI fusion is often difficult to calculate.In order to improve the computational efficiency of multi-sensor fusion,the Gaussian components are clustered and then isolated using distance metric in this paper.The distance metric can calculate the density weights of the targets after fusion,and by discarding the fusion assumptions that the weights can be ignored,a simplified approximate density function can be constructed.Analysis shows that the proposed fusion algorithm can achieve a multiple improvement in computational efficiency compared to traditional GCI fusion algorithms.The effectiveness of the proposed fusion algorithm is verified through experiments in a simulation scenario where 12 targets appeared successively.