天波雷达RD图干扰检测器的图库设计与纹理特征研究  

Image Dataset Construction and Textural Feature Analysis for Interference Detector Design Based on RD Images in Sky-wave OTH Radar

作　　者：郑园园[1] 罗忠涛 张安安 王涛[2] 马超[1] ZHENG Yuanyuan;LUO Zhongtao;ZHANG Anan;WANG Tao;MA Chao(Nanjing Research Institute of Electronics Technology,Nanjing Jiangsu 210039,China;School of Communication and Information Engineering,Chongqing University of Posts and Telecommunications,Chongqing 400065,China)

机构地区：[1]南京电子技术研究所,江苏南京210039 [2]重庆邮电大学通信与信息工程学院,重庆400065

出　　处：《现代雷达》2023年第7期9-14,共6页Modern Radar

摘　　要：讨论天波雷达基于距离-多普勒(RD)图像的射频干扰与瞬态干扰检测新方法,研究其关键技术即RD图库设计和纹理特征,以支撑图像分类器的有效设计。首先提出RD图分类器的设计思路,介绍图库建设方法;然后甄选3种纹理特征算子,提出双纹理特征融合方法,强化弱干扰检测能力;最后,基于K近邻算法设计分类器,以实测数据作测试图库,结果表明对强干扰检测准确率达到98%以上,对弱瞬态和弱射频干扰的检测准确率达到75%和85%。因此,本文图库设计和纹理特征可以有效支撑RD图干扰检测。This paper discusses the novel method of transient interference and radio frequency interference detection based on range-Doppler(RD)images for sky-wave over-the-horizon(OTH)radar,and emphasizes on the key techniques on RD image dataset construction and textural feature extraction,to support the effective design of RD image classifiers.Firstly,the routine of RD image classifier design is proposed and the method of dataset construction is introduced.Then,three operators for the RD image textural feature extraction and feature fusion algorithms are analyzed to improve the performance of weak interference detection.Finally,the classifiers are designed based on the K-nearest neighbor algorithm and employs the real image dataset as the test set.The classifiers can achieve the detection probability over 98%for strong interference and the detection probabilities 75%and 85%for weak transient interference and weak radio frequency interference,respectively.Hence,our method of image dataset construction and textural feature design is effective for supporting the RD image interference detection.

关 键 词：天波超视距雷达 干扰检测 图像分类 纹理特征 特征融合 

分 类 号：TN958.93[电子电信—信号与信息处理]