提出一种新的基于广义交叉协方差(GCI)和平均边缘概率密度(AMD)的分布式多目标融合机制。传统的多传感器跟踪算法存在一些难以克服的缺点: 如航迹关联性能对关**数设置敏感, 航迹关联计算量随目标个数增加呈指数增加等。为解决这些问题, 提出一种适用于多目标的鲁棒性较好的分布式融合算法。首先, 将多目标联合后验概率密度近似为关于其AMD的积分布; 然后, 考虑到不同雷达节点之间的未知的互相关, 采用GCI融合算法进行融合。由于航迹关联过程被内嵌在航迹融合过程中, 所提分布式融合机制同时进行航迹关联和航迹融合, 理论上会更加完备。所提算法借助高斯混合模型实现, 试验结果证明其性能相比传统融合算法有很大改善。

This paper proposes a novel distributed multi-target fusion mechanism based on Generalized Covariance Intersection (GCI) and Average Marginal Density (AMD). There are several inevitable defects in traditional multi-sensor tracking methods. For instance, the track-correlation performance is sensitive to the setting of correlation parameters, and the calculated amount of track correlation increases exponentially with the number of targets, etc. To solve these problems, a robust distributed fusion method suitable for multiple targets is proposed. Firstly, we approximated the local multi-target posterior probability density as a product distribution with its AMD. Secondly, considering the unknown correlation between different radar nodes, the GCI fusion algorithm was employed to perform distributed fusion. Since the track correlation process was embedded in the track fusion process, the distributed fusion mechanism performed track correlation and track fusion at the same time. Finally, we derived the closed-form solution of GCI fusion with AMDs. The proposed fusion algorithm is implemented by using Gaussian mixture model, and the experimental result shows that its performance is superior to that of the traditional fusion algorithm.