Design on temporal-spatial Transformer model for air target intention recognition

面向空中目标意图识别的时空Transformer模型设计

Air and Missile Defense College, Air Force Engineering University, Xi'an 710051, China

Received: 30 June 2023

Abstract

The battlefield changes rapidly in information warfare. The battlefield situation data presents massive and diversified characteristics, which makes it increasingly difficult to identify the operational intent of air targets based on expert experience. In combination with the current state-of-the-art intelligent methods, the Transformer model is studied and introduced into the field of air target intent recognition for the first time, and a new intent recognition method temporal-spatial transformer(TST) is designed, which can effectively mine the deep feature information in the temporal and spatial domains of battlefield situational data to improve the accuracy of air target combat intention recognition. In addition, a comparative study of the four currently advanced neural network intention recognition methods shows that TST achieved outstanding performance in all indicators and outperformed all compared neural network models. TST method not only has excellent accuracy but also extremely fast convergence rate, which allows it to quickly capture key information from battlefield situation data for intention recognition.

摘要

信息化条件下的战争环境瞬息万变, 战场态势数据呈现海量、多样化的特点, 导致利用专家经验识别空中目标作战意图的难度越来越高。结合目前先进的智能化方法, 对Transformer模型进行研究并将其引入空中目标意图识别领域, 设计出一种新的意图识别方法Temporal-Spatial Transformer(TST), 可以有效地挖掘战场态势数据中时间域和空间域的深层特征信息, 提高空中目标作战意图识别准确率。同时, 对4种目前较为先进的神经网络意图识别方法进行效果对比, 结果表明, TST模型在各类指标上都取得了突出的效果, 优于所有对比的神经网络模型。TST模型不仅有优异的准确率, 而且收敛速度极快, 可以迅速地抓取战场态势数据中的关键信息进行意图识别。