基于LSTM的无人机轨迹识别技术研究  

Research on UAV Trajectory Recognition Based on LSTM

作　　者：杨星鑫 吕泽均[1] YANG Xing-xin;LV Ze-jun(National Key Laboratory of Air Traffic Control Automation System Technology,College of Computer Science,Sichuan University,Chengdu 610065)

机构地区：[1]四川大学计算机学院,国家空管自动化系统技术重点实验室,成都610065

出　　处：《现代计算机》2020年第5期18-22,25,共6页Modern Computer

摘　　要：无人机目标逐渐成为威胁低空空域安全的重要因素,针对低空空域对目标识别的实时性和准确性需求,提出基于目标轨迹进行无人机识别的方法。构造目标飞行的轨迹特征,使用雷达探测到的目标飞行数据训练模型,对无人机目标进行识别。实验结果表明,使用递归神经网络(RNN)的长短期记忆(LSTM)结构建模分析目标的轨迹信息,具有精度高、易实现的特点。Drone targets are becoming an import threat to the safety of low-altitude airspace.In view of the real-time and accuracy requirements of low-altitude airspace for target identification,amethodfor UAV recognition based on target trajectory is proposed.Construct the trajectory characteristics of the target flight,use the radar to detect the target flight data training model,and identify the target of the UAV.The experimental results show that the trajectory information of the target is highly accurate and easy to implement by using the Long Short Term Memory structure modeling of the recursive neural network.

关 键 词：无人机 目标识别 轨迹 递归神经网络 长短期记忆 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] TP391.41[自动化与计算机技术—控制科学与工程] V279[航空宇航科学与技术—飞行器设计]