无人机雷达航迹运动特征提取及组合分类方法  

Motion feature extraction and ensembled classification method based on radar tracks for drones

作　　者：刘佳[1] 徐群玉 陈唯实 LIU Jia;XU Qunyu;CHEN Weishi(Research Institute for Frontier Science,Beihang University,Beijing 100191,China;Research Institute of Civil Aviation Law,China Academy of Civil Aviation Science and Technology,Beijing 100028,China;Airport Research Institute,China Academy of Civil Aviation Science and Technology,Beijing 100028,China)

机构地区：[1]北京航空航天大学前沿科学技术创新研究院,北京100191 [2]中国民航科学技术研究院民航法规与标准化研究所,北京100028 [3]中国民航科学技术研究院机场所,北京100028

出　　处：《系统工程与电子技术》2023年第10期3122-3131,共10页Systems Engineering and Electronics

基　　金：国家自然科学基金委员会中国民航局民航联合研究基金(U1933135,U1633122);北航卓越百人计划双一流引导专项(ZG216S2182)资助课题。

摘　　要：飞鸟和无人机目标的雷达回波存在高度相似性,区分难度较大。因此,对无人机、飞鸟以及动态降水杂波形成的目标航迹的时空间特征进行了研究,分析了无人机和飞鸟在运动机理以及行为模式上的差异,提出了一种基于目标航迹的运动特征提取方法,并构建了目标特征向量。基于探鸟雷达系统提供的目标实测航迹数据,建立了训练和测试样本集,采用监督类学习方法并结合随机森林模型实现了对无人机、飞鸟和降水杂波目标航迹的区分。实验结果表明,在广域范围内,无人机目标的正确识别率可达85%以上,分类器模型的运算效率高,样本适应性强,具备较好的普适性和实用价值。The radar echoes of birds and drones target have high similarity,which make it difficult to distinguish them.Therefore,the spatio-temporal characteristics of target tracks formed by drones,birds and dynamic precipitation clutter are studied,and the differences in motion mechanisms and behavior patterns between drones and birds are analyzed.A motion feature extraction method based on target tracks is proposed and target feature vectors are constructed.Based on the measured track data of the target provided by the detection bird radar system,a training and test sample set is established.The supervised learning method combined with the random forest model is used to distinguish the target tracks of drones,birds and precipitation clutter.The experimental results show that the correct recognition rate of drone targets over a wide area can reach over 85%,and the classifier model has high calculation efficiency,strong sample adaptability,and good universality and practical value.

关 键 词：无人机检测 雷达目标识别 特征提取 监督类学习 

分 类 号：TN959.1[电子电信—信号与信息处理]