单目标一维成像及特征提取方法性能分析  

One-Dimensional Imaging of Single Target and Performance Analysis of Target Feature Extraction Methods

作　　者：沈姗姗[1] 陆锦辉[2] 顾苏[3] 

机构地区：[1]南京理工大学紫金学院,江苏南京210046 [2]南京理工大学,江苏南京210094 [3]南京船舶雷达研究所,江苏南京210003

出　　处：《雷达科学与技术》2011年第5期420-424,共5页Radar Science and Technology

摘　　要：为后续目标识别作准备,针对采集并且存储之后的飞机目标雷达回波信号,完成后端雷达信号处理。首先,基于MATLAB提取波门内采样的回波信号并且实现目标的一维成像。然后,采用传统的傅里叶变换模值、双谱奇异值分解法对目标的一维距离像进行特征提取。为了获得较为稳定的目标特征,针对以上两种方法,主要从特征相似性的测量角度分析单一目标特征提取方法的性能优劣。最后通过实验,从相关系数概率密度分布情况得出双谱奇异值分解特征提取法性能较佳,所得到的目标特征较稳定。In this paper, in order to prepare for the follow-up target recognition, after sampling and storage of aircraft radar echo signal, radar back-end signal processing is completed. First of all, the echo sampling in wave gate is extracted and one dimensional range profiles of the target is fulfilled based on MATLAB. After that, the traditional modulus of Fourier transform and bispectrum singular value decomposition are used to achieve feature extraction of one dimensional range profiles of the target. In order to obtain relatively stable feature of the target, mainly from the point of similarity measurement, advantages and disadvantages of the performance of single target feature extraction are analyzed according to the two feature extraction methods. At last, through the experiment, the probability density distribution of similarity shows that performance of feature extraction method of bispectrum singular value decomposition is much better and the feature is much more stable.

关 键 词：特征提取 一维成像 双谱奇异值分解 相似性 

分 类 号：TN957[电子电信—信号与信息处理]