基于广义概率假设密度的多目标运动估计方法研究

Research on Multi-Target Motion Estimation Method Based on Generalized Probability Hypothesis Density

面向空对空作业的发展需求,提出了一种基于广义概率假设密度的多目标运动估计方法。在Faster-RCNN方法基础上引入多尺度分析,并利用改进K-means方法对观测目标进行粗聚类,以此为前置输入,提出了基于广义泊松分布的概率假设密度滤波器,将聚类信息纳入滤波估计的权重更新中,增强了对变阵群目标的跟踪时效性。仿真结果表明,本文方法在没有初始聚类信息的先验知识下,依然能够完成对多目标的识别分类与跟踪,且精度优于现有的集群目标运动估计方法。

Aiming at the development needs of air-to-air operations,this paper proposes a multi-target motion estimation method based on generalized probability hypothesis density.Multi-scale analysis is introduced into based on the Faster-RCNN algorithm,and the improved K-means method is used to perform coarse clustering on the observed targets.With this as a pre-input,a probability hypothesis density filter based on generalized poisson distribution is proposed,and the clustering information is included in the weight update of the filter estimate to enhance the tracking timeliness of targets for the variable group.The simulation results show that the proposed method can still complete the recognition and classification of multiple targets without prior knowledge of the initial clustering information,and is superior to present swarm target motion estimation method in precision.