基于时频变换的弹道目标微多普勒特征提取  

Ballistic Target Micro-Doppler Feature Extration Method Based on the Time-frequency Analysis

作　　者：田巳睿[1] 蒋跃红[1] 郭汝江[1] 施西野[1] 

机构地区：[1]南京电子技术研究所,南京210039

出　　处：《现代雷达》2012年第1期40-43,48,共5页Modern Radar

摘　　要：微多普勒特征是识别弹道目标的重要手段。文中从空间弹道导弹和诱饵散射点微动模型出发,提出了一种基于时频二维像不变矩的微多普勒特征提取方法。该方法首先对导弹和诱饵进行时频变换,然后提取了图像中的7个Hu不变矩作为目标识别的特征。通过仿真实验,分析了特征组合对识别效果的影响,并提出了最优特征组合。文中还评估了特征矢量在不同信噪比条件下的稳定性,试验结果表明,时频像不变矩特征对信噪比变化具有一定的稳定性。According to the mieo-motion models of the ballistic targets, a micro-Doppler feature extraction method based on time-frequency analysis and moment invariant is proposed in this paper. In this approach, time-frequency analysis is applied at the very beginning. Then the Hu invariant moments are extracted from the time-frequency imagery. Simulations of precession warhead and swing bait are carried out, whereafter feature combinations are analyzed and the optimized feature vector is achieved. Furthermore, the feature robustness with different SNR is evaluated.

关 键 词：微动目标识别 平滑伪Wigner-Ville分布 不变矩 

分 类 号：TJ760[兵器科学与技术—武器系统与运用工程]