基于GTN与证据理论的集群目标意图分层识别方法  

A hierarchical identification method of cluster target intent based on gated-transformer-network and evidence theory

作　　者：李泽鹏 吉琳娜[1,2] 杨风暴 LI Zepeng;JI Linna;YANG Fengbao(School of Information and Communication Engineering,North University of China,Taiyuan 030051,China;Key Laboratory of Intelligent Information Control Technology of Shanxi Province,Taiyuan 030051,China)

机构地区：[1]中北大学信息与通信工程学院,太原030051 [2]智能信息控制技术山西省重点实验室,太原030051

出　　处：《国外电子测量技术》2025年第1期22-29,共8页Foreign Electronic Measurement Technology

基　　金：山西省基础研究计划项目(202203021221104)。

摘　　要：针对现有集群目标意图识别方法中意图特征仅从整体特征构建、忽略集群目标协同作战的聚合性而导致识别准确率低的问题,提出了一种集群意图的分层识别方法。首先对集群目标成员的数值型特征进行初级识别,然后考虑集群目标的通信协同以及属性信息,将单目标的意图结果转化为各自对集群意图的支持向量。最后对集群意图支持向量进行证据融合,得到最终的集群意图结果。实验结果表明,与现有方法相比,所提方法在准确率、精确率、召回率以及F1指数上表现最优,且与整体特征的识别方法相比,准确率提升了7.4%,证明了所提方法的有效性。The intention features of the existing swarm target intent recognition methods are all constructed from the overall features,and the recognition accuracy is low due to ignoring the aggregation of swarm target cooperative operations.This paper proposes a hierarchical identification method for cluster intent.Firstly,the numerical features of the cluster target members were initially identified,and then the single-target intent result was transformed into their respective support vectors for the cluster intent considering the communication coordination and attribute information of the cluster target.Finally,the evidence fusion of the cluster intent support vector was carried out to obtain the final cluster intent result.Experimental results show that the proposed method has the best performance in terms of accuracy,precision,recall and F1 index compared with the existing methods,and the accuracy is increased by 7.4%compared with the overall feature recognition method,which illustrates the effectiveness of the method.

关 键 词：集群目标 意图识别 分层 支持向量 证据融合 

分 类 号：TP173[自动化与计算机技术—控制理论与控制工程]