基于RMDLPP的雷达空中目标分类  

Classification of radar air targets based on RMDLPP

作　　者：刘帅康 曹伟 管志强 杨学岭 许金鑫 LIU Shuaikang;CAO Wei;GUAN Zhiqiang;YANG Xueling;XU Jinxin(The 724th Research Institute of China Shipbuilding Group Corporation,Nanjing 211153,China;Colledge of Electronic and Information Engineering,Nanjing University of Aeronautics and Astronautics,Nanjing 211106,China)

机构地区：[1]中国船舶集团有限公司第七二四研究所,江苏南京211153 [2]南京航空航天大学电子信息工程学院,江苏南京211106

出　　处：《系统工程与电子技术》2024年第4期1220-1228,共9页Systems Engineering and Electronics

摘　　要：针对鉴别局部保持投影(discriminant locality preserving projections, DLPP)在窄带雷达目标数据降维时出现的类内离散度矩阵奇异和对孤立点敏感进而导致类别之间可分性弱的问题,提出了基于鲁棒性边界DLPP(robust margin DLPP, RMDLPP)的窄带雷达空中目标分类方法。首先,在计算样本之间距离时将两样本点的欧氏距离与同类样本均值相关联。然后,挑选一定数量的边界样本点进行处理并对优化DLPP目标函数进行降维。最后,使用高性能分类器对降维后的数据进行区分,实现对空中目标的分类。通过对X波段对空警戒雷达实测数据的对比实验表明,所提方法具有更好的分类准确率和对孤立点的鲁棒性。Aiming at the problem of exoticism of the intraclass dispersion matrix and sensitivity to isolated points in narrowband radar target data reduction of discriminant locality preserving projections(DLPP)in narrow-band radar target data,a narrow-band radar air targets classification method based on robust margin DLPP(RMDLPP)is proposed.Firstly,the Euclidean distance of the two sample points is correlated with the homogeneous sample mean value when calculating the distance between samples.Then,a certain number of boundary sample points are selected for processing and the DLPP objective function is optimized for dimensionality reduction.Finally,a high-performance classifier is used to distinguish the dimensionality reduction data and achieve the classification of aerial targets.Comparative experiments on X-band air-to-air alert radar measurements show that the proposed method has better classification accuracy and robustness to isolated points.

关 键 词：窄带雷达 空中目标分类 鉴别局部保持投影 最大边界准则 降维 

分 类 号：TN957[电子电信—信号与信息处理]