基于训练特征空间分布的雷达地面目标鉴别器设计

发布时间：2018-01-25 19:00

本文关键词： 目标鉴别高分辨距离像 K-Means聚类支撑向量域描述 K近邻分类器　出处：《电子与信息学报》2016年04期 　论文类型：期刊论文

【摘要】：该文对雷达地面目标高分辨1维距离像目标识别中的库外目标鉴别问题,提出一种基于训练特征空间分布的雷达地面目标鉴别器。在训练阶段利用基于相关系数预处理的K-Means聚类方法对库内目标样本特征空间进行区域划分,并采用基于空间分布的支撑向量域描述方法确定样本特征空间的边界与支撑向量,利用样本特征空间边界与加权K近邻原则对目标类别进行判决。该方法解决了库内目标与库外目标的鉴别问题,提高了目标识别系统的总体性能。针对多种不同姿态下目标特征空间非均匀聚合的特点,对训练样本特征空间进行区域划分,减小模板匹配搜索运算规模,保证目标鉴别所需的实时性工作要求。最后通过仿真和实测数据验证了该方法具备优良的鉴别性能与良好的实时处理能力。
[Abstract]:This paper deals with the problem of target identification outside the database in radar ground target recognition with high resolution 1 D range profile. In this paper, a radar ground target discriminator based on training feature space distribution is proposed. In the training stage, K-Means clustering method based on correlation coefficient preprocessing is used to partition the target sample feature space in the database. . The support vector domain description method based on spatial distribution is used to determine the boundary and support vector of the sample feature space. By using the boundary of sample feature space and the weighted K-nearest neighbor principle, the classification of target is determined. This method solves the problem of identifying the target in and out of the library. The overall performance of the target recognition system is improved. Aiming at the non-uniform aggregation of target feature space under different attitude, the training sample feature space is divided to reduce the scale of template matching search. Finally, the simulation and real time data show that the method has good discriminant performance and good real-time processing capability.
【作者单位】：西安电子科技大学雷达信号处理国家重点实验室;
【分类号】：TN957.52
【正文快照】： 1引言弹载雷达地面目标识别系统主要面临的问题是复杂环境下对特定目标的实时鉴别。针对弹载雷达应用背景,采用合成宽带雷达体制生成雷达地面目标1维高分辨距离像(High Resolution RangeProfile,HRRP)用于检测与识别。目标HRRP反映了目标散射点在雷达视线方向的几何结构信息,

【相似文献】