基于注意力机制的BiGRU战场目标意图判别  

BiGRU Battlefield Target Intention Discrimination Based on Attention Mechanism

作　　者：王姝佳 肖秦琨[1] 华瑾[1] 贾松涛 WANG Shu-jia;XIAO Qin-kun;HUA Jin;JIA Song-tao(Xi’an Technological University,Xi’an Shaanxi 710021,China)

机构地区：[1]西安工业大学,陕西西安710021

出　　处：《计算机仿真》2024年第4期27-32,共6页Computer Simulation

基　　金：国家自然科学基金(62071366);陕西省自然科学基金(2020JM566);西安市智能兵器重点实验室(2019220514SYS020CG042)。

摘　　要：针对现代战场环境信息复杂化、多样化,战场目标具有隐蔽性、欺骗性等各种问题,提出一种基于双向门控循环单元(BiGRU)的战场目标意图识别方法。为了使网络更好的关注信息重点,提高目标意图识别的准确率,在BiGRU中加入了注意力机制,注意力机制可以使模型更加关注与预测结果有紧密联系的部分,忽略相关性不高的地方,同时采用Zoneout机制来减少网络过拟合。实验结果得出,结合注意力机制的BiGRU战场目标意图判别方法准确率高达97.1%,比基础GRU算法准确率提高12%。Aiming at the complexity and diversity of modern battlefield environment information and the hidden and deceptive battlefield targets,a battlefield target intention recognition method based on bidirectional gated cyclic unit(BiGRU)was proposed.In order to make the network pay more attention to the key points of information and improve the accuracy of target intention recognition,an attention mechanism was added into BiGRU.The attention mechanism can make the model pay more attention to the parts closely related to the prediction results and ignore the places with low correlation.Moreover,Zoneout mechanism was adopted to reduce the network overfitting.The experimental results show that the accuracy of BiGRU method combined with attention mechanism is up to 97.1%,which is 12%higher than that of the basic GRU algorithm.

关 键 词：注意力机制 战场目标 意图判别 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]