基于Transformer网络的无人机目标航迹关联方法  

A UAV track correlation method based on Transformer network

作　　者：张强[1,2] 王淳 陈亚伟 ZHANG Qiang;WANG Chun;CHEN Yawei(Nanjing Research Institute of Electronics Technology,Nanjing 210039,China;National Key Laboratory of Radar Detection and Perception,Nanjing 210039,China)

机构地区：[1]南京电子技术研究所,南京210039 [2]雷达探测感知全国重点实验室,南京210039

出　　处：《空天预警研究学报》2025年第2期90-92,98,共4页JOURNAL OF AIR & SPACE EARLY WARNING RESEARCH

基　　金：国家自然科学基金资助项目(U19B2031)。

摘　　要：针对传统航迹关联方法在密集目标场景下存在关联错误率高、关联效率低等问题,本文提出了一种基于Transformer的航迹关联方法.该方法通过Transformer网络自动抽取关联特征,自动学习航迹关联准则,实现了对雷达航迹端到端关联.仿真结果表明,与传统的联合概率数据关联(JPDA)方法相比,本文提出的方法具备较好的航迹关联能力.该方法为解决密集目标场景下航迹关联提供了一种解决思路.In order to solve the problems of high correlation error rate and low correlation efficiency of traditional trajectory correlation methods in dense target scenarios,this paper proposes a Transformer network-based trajectory correlation method,in which the Transformer network is used to automatically extract correlation features and automatically learn trajectory correlation criteria,with the end-to-end correlation of radar trajectories achieved.The simulation results show that compared with the traditional joint probability data association(JPDA)method,the method proposed in this paper has better trajectory correlation ability.This method provides a solution for trajectory correlation in dense target scenarios.

关 键 词：无人蜂群 航迹关联 Transformer网络 端到端方法 

分 类 号：TN957.51[电子电信—信号与信息处理]