基于多角度合成SAR图像的目标识别性能分析

Target recognition performance analysis based on multi-aspect composite SAR images

合成孔径雷达(synthetic aperture radar,SAR)在合成孔径累积时间内仅在小范围方位角获取目标的后向散射特性,导致SAR图像对观测方位向的变化极其敏感。通过图像非相干合成方法将不同方位向上的多幅同目标SAR图像合成单幅特征更明显的SAR图像,通过二维主成分分析方法提取特征和k-近邻分类方法实现目标识别。在两组不同数据集上对识别性能进行分析。实验结果表明,多角度SAR的识别率比单一角度更高。多角度SAR对观测平台俯视角的变化具有较强的鲁棒性。

Synthetic aperture radar(SAR)receives the backscatter of target within only a small range of azimuth during the synthetic aperture accumulation time,which makes the SAR image extremely sensitive to the change in observation azimuth.In this study,an SAR image with more obvious features is composed of multiple SAR images obtained at different azimuths by means of non-coherent composition,and then 2 DPCA is used to extract features and the k-nearest neighbor method is used for target recognition.Finally the recognition performance is analyzed on two different datasets.The experimental results show that the multi-aspect SAR has a higher recognition rate than the single aspect,and has strong robustness to the change in depression angle.