论文部分内容阅读
人造目标检测是极化合成孔径雷达(synthetic aperture radar,SAR)图像自动解译中的重要环节。该文提出了一种基于Riemann核Fisher准则的人造目标检测方法。核函数通过Hermite正定矩阵流形上的Riemann度量来构造。极化协方差矩阵映射到核函数诱导的高维特征空间后用Fisher准则进行判别。该方法考虑到了极化SAR数据特殊的矩阵结构,并且不需要任何统计模型假设,因而特别适于检测极化SAR图像中的人造目标。以舰船目标检测为应用背景验证了该方法的有效性。实验结果表明:该方法优于其他常用的检测器,特别是在低目标杂波比条件下。
Man-made target detection is an important part in the automatic interpretation of synthetic aperture radar (SAR) images. In this paper, an artificial target detection method based on Riemann kernel Fisher criterion is proposed. The kernel function is constructed by the Riemann metric on Hermite positive definite matrix manifolds. Polarization covariance matrix is mapped to kernel-induced high-dimensional feature space and discriminated by Fisher criterion. This method takes into account the special matrix structure of polarized SAR data and does not require any statistical model assumptions and is therefore particularly suitable for detecting man-made objects in polarized SAR images. The effectiveness of this method is verified by the detection of ship target. Experimental results show that this method is superior to other commonly used detectors, especially in the low target clutter ratio conditions.