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。
Detection of man-made targets is essential for automatic interpretation of polarimetric synthetic aperture radar (SAR) images. This paper describes a man-made target detection method that utilizes the Riemannian kernel Fisher criterion. The kernel function is constructed by means of a Riemannian metric defined on the manifold of Hermitian positive definite matrices. The polarimetric covariance matrices are mapped into the high-dimensional feature space induced by the kernel function and then discriminated by the Fisher criterion. This method takes into account the special matrix structure of the polarimetric SAR data and does not assume any statistical models; therefore, it is particularly suitable for detecting man-made targets in polarimetric SAR images. The effectiveness of this method is verified in the context of ship detection. Tests show that this method outperforms other detectors, especially for low target-to-clutter ratio.
[1] Novak L M, Sechtin M B, Cardullo M J. Studies of target detection algorithms that use polarimetric radar data [J]. IEEE Trans Aerosp Electron Syst, 1989, 25(2): 150-165.
[2] Novak L M, Burl M C. Optimal speckle reduction in polarimetric SAR imagery [J]. IEEE Trans Aerosp Electron Syst, 1990, 26(2): 293-305.
[3] Yang J, Yamaguchi Y, Boerner W M, et al. Numerical methods for solving the optimal problem of contrast enhancement [J]. IEEE Trans Geosci Remote Sens, 2000, 38(2): 965-971.
[4] Chaney R D, Bud M C, Novak L M. On the performace of polarimetric target detection algorithms [J]. IEEE Aerosp Electron Syst Mag, 1990, 5(11):10-15.
[5] D'Hondt O, Guillaso S, Hellwich O. Iterative bilateral filtering of polarimetric SAR data [J]. IEEE J Sel Topics Appl Earth Observ Remote Sens, 2013, 6(3): 1628-1639.
[6] Song H, Yang W, Xu X, et al. Unsupervised PolSAR imagery classification based on Jensen-Bregman LogDet divergence [C]//Proceedings of the 10th European Conference on Synthetic Aperture Radar. Berlin, Germany: VDE, 2014: 915-918.
[7] Song H, Yang W, Bai Y, et al. Unsupervised classification of polarimetric SAR imagery using large-scale spectral clustering with spatial constraints [J]. Int J Remote Sens, 2015, 36(11): 2816-2830.
[8] Wang Y, Han C. PolSAR image segmentation by mean shift clustering in the tensor space [J]. Acta Autom Sin, 2010, 36(6): 798-806.
[9] Yang F, Gao W, Xu B, et al. Multi-frequency polarimetric SAR classification based on Riemannian manifold and simultaneous sparse representation [J]. Remote Sens, 2015, 7(7): 8469-8488.
[10] Sra S. Positive definite matrices and the S-divergence [J]. Proc Amer Math Soc, 2015.
[11] Jayasumana S, Hartley R, Salzmann M, et al. Kernel methods on Riemannian manifolds with Gaussian RBF kernels [J]. IEEE Trans Pattern Anal Mach Intell, 2015, 37(12): 2464-2477.
[12] Shawe-Taylor J, Cristianini N. Kernel Methods for Pattern Analysis [M]. Cambridge, UK: Cambridge University Press, 2004.
