基于元度量学习的小样本空战目标意图识别方法
A Small Sample Air Combat Target Intent Recognition Method Based on Meta-Metric Learning
张灏龙 1权晓伟 1刘瑞峰 1黎开颜1
作者信息
- 1. 中国航天系统科学与工程研究院,北京 100048
- 折叠
摘要
针对战场复杂环境下通过较少空战对抗数据识别作战意图的问题,提出基于元度量学习框架的作战意图识别方法.该方法通过构建基于双向门控循环单元网络,实现对空战时序数据的有效特征提取,进而提出注意力机制,促使网络实现对小样本空战数据时序核心特征的充分提取,从而获取类间差异,达到较高的空战意图识别的准确率和速度.仿真实验表明,所提方法对于空战目标意图识别具有较好的准确率和实时性,在小样本数据的情况下能够实现较好的识别性能.
Abstract
In order to achieve rapid and accurate recognition of combat intent in complex battlefield envi-ronments with fewer air combat data samples,a combat intent recognition model based on meta-metric learning framework is proposed.Regarding received small sample data during air combat intent recognition,a two-way gated recurrent unit network based on the air combat temporal data are developed to realize effec-tive feature extraction,and the attention mechanism is introduced for promoting the network to fully extract the temporal core features of the air combat data when facing the small sample data are merely available for obtaining the inter-class differences which finally achieve a fairly higher recognition accuracy and recogni-tion speed.Simulation results show that the model proposed in this paper has better accuracy and real-time performance for air combat target intent recognition,especially in the case of small sample data.
关键词
空战目标/意图识别/注意力机制/双向门控循环单元网络/元度量学习Key words
Air combat targets/Intent recognition/Attentional mechanism/Bi-directional gated recurrent cell networks/Meta-metric learning引用本文复制引用
出版年
2024
NETL
NSTL
万方数据