复杂环境下少样本域自适应雷达RD智能识别技术  

Radar RD Intelligent Recognition Based on Domain Adaption with Tiny Data in Complex Environments

作　　者：石晓荣[1] 张康[1] 倪亮[1] 刘泽文[1] 姜丰 陈鑫 Shi Xiaorong;Zhang Kang;Ni Liang;Liu Zewen;Jiang Feng;Chen Xin(Beijing Institute of Control and Electronics Technology,Beijing 100038,China)

机构地区：[1]北京控制与电子技术研究所,北京100038

出　　处：《航天控制》2023年第4期20-26,共7页Aerospace Control

摘　　要：针对复杂电磁环境下,传统基于雷达回波的目标识别方法定位精度差和识别能力不足的问题,以典型海背景环境下的雷达时频二维像目标为研究对象,提出了复杂环境下少样本域自适应雷达智能RD识别技术,通过设计基于分离注意力的偏移区间配准目标识别算法架构,实现了对目标的精确定位和识别,同时设计了域自适应技术对源域和目标域进行特征配准,解决少样本条件下算法场景适应性不足的问题。通过在雷达回波仿真数据和真实数据上对算法测试验证,取得了较优的结果,证明了该算法的有效性。Aiming at the low detection accuracy and weak recognition ability of traditional radar object detection and recognition in the wild,the radar echo is token in the typical ocean environment as the research object in this paper,and a domain adaptive radar RD intelligent recognition technology with tiny data is proposed in complex environments.A two-stage detector is proposed,which is able to acquire perfect performance through this research simulation,and the domain adaptive algorithm designed for the source domain and the target domain can provide solution of lacking generalization adaptability under tiny quantity data.The model shows a good performance result during training and testing by the simulation data and the real data,which demonstrats the validation and capability.

关 键 词：雷达RD智能识别 少样本 域自适应 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]