一种子像素精度SAR图像目标峰值提取方法

A Method of Target's Peak Extraction from SAR Imagery in Sub-pixel Accuracy

SAR图像目标峰值是SAR图像目标识别的重要特征 ,它本质对应于目标散射中心 ,目标峰值提取是SAR图像目标识别的一个重要步骤。基于峰值SAR图像目标识别系统要求目标峰值提取方法应具有快速、高精度的特点。现有的SAR图像目标峰值提取方法精度较低 ,其对目标峰值位置的估计精度只能达到像素级。为了提高SAR图像目标峰值提取精度 ,该文在分析SAR图像峰值模型基础上 ,提出了一种子像素精度SAR图像目标峰值提取方法 ,并通过仿真实验 ,分析了该方法对目标峰值位置、幅度的估计性能 ,结果表明在SNR =2 0dB的情况下 ,该方法对目标峰值位置估计的标准偏差 <0 .1个像素 ,峰值幅度估计的标准偏差小于 0 .0 5 H(这里H表示目标峰值的真实幅度 ) ,文中还给出了该方法对实测MSTARSAR图像的目标峰值提取结果。

Target's peak is an important feature for SAR image target recognition, which corresponds to target's scatter center in essence. Accordingly, peak extraction is an important process in SAR image target recognition. To meet the peak-based SAR image target recognition system's need, the method of target's peak extraction should be fast and accurately enough. But, the accuracy of the existing method of target peak extraction is comparatively low and the estimation of peak's position can only reach pixel-accuracy. In order to improve the accuracy of target's peak extraction, based on the analysis of SAR imagery's peak model, a method of target's peak extraction in sub-pixel accuracy is proposed in this paper and the estimate performance of peak's position and amplitude is analyzed through simulation. The results show that for SNR=20dB, the standard bias of peak's position estimation is less than 0.1 pixel, and the standard bias of peak's amplitude estimation is less than 0.05*H (Here, H denotes the true amplitude of simulated peak), also, the result of target's peak extraction for real MSTAR SAR imagery is given.