摘要
战场态势意图由一系列战术动机组成,具有时序、动态、多目标等特点。现有的态势感知方法中存在只研究单目标或忽略时序性的问题。在海战多目标、长周期的背景下,针对以上问题,提出了一种基于ResNet-ViT网络的海战综合态势感知模型。其中ResNet(Residual Network)网络用于提取目标之间的空间特征,而ViT(Vision Transformer)网络则利用Transformer能够挖掘长距离依赖的特性来捕获时序特征。结果表明:模型能以92%~95%的准确率(预测正确的样本的数量与总样本数量的比例)预测海战意图,解决了长周期下多目标协同作战的意图预测问题。
Battlefield situation intent consists of a series of tactical motivations,which are time series,dynamic and multi-objective.However,the existing situational awareness methods have the problem of only studying single objectives or ignoring time series.Based on the characteristics of the multiple-objective and long period in the background of naval warfare,a comprehensive situational awareness model of naval warfare based on the ResNet-ViT network is presented.The residual neural network(ResNet)extracts spatial features between targets,whereas the vision transformer(ViT)network captures time-series features by using the transformer's ability to mine long-distance dependency features.The experimental results show that the proposed model can predict the intent of naval warfare with 92%~95%accuracy(proportion of correctly predicted samples to total samples),which solves the intention prediction problem of multi-target cooperative warfare over a long period.
作者
朱小勇
陈胜
ZHU Xiaoyong;CHEN Sheng(School of Optical-Electrical and Computer Engineering,University of Shanghai for Science and Technology,Shanghai 200093,China)
出处
《信息与控制》
CSCD
北大核心
2023年第5期638-647,共10页
Information and Control
基金
国家自然科学基金(81101116)
关键词
海战
态势感知
深度学习
多目标
时序特征
naval battle
situational awareness
deep learning
multi-objective
time series feature
作者简介
朱小勇(1994-),男,硕士生。研究领域为图像处理与分析,态势感知,深度学习等;通信作者:陈胜(1976-),男,博士,副教授。研究领域为图像处理与分析,计算机辅助诊断,态势感知等,chnshn@hotmail.com。
