雷达极化对角加载检测器的最优权重算法

The Algorithms for Optimal Weights of Polarimetric Diagonal Loading Detector

针对复杂海杂波环境下极化SAR图像弱小目标检测难题,最近提出的雷达极化对角加载滤波器融合了极化白化滤波器(PWF)和极化检测优化滤波器(PDOF)的优势,通过选择合适的线性加权系数,能够突破非高斯杂波背景下的目标检测性能限制。但是如何获取该线性组合的最优加权权重是该方法的难点。首先重新构造了基于线性组合的雷达极化对角加载检测器,在此基础上从曲线下面积(AUC)的检测性能评估角度给出了线性组合的最优化数学模型,并提出了基于该准则的最优权重算法。为加快求解速度,分别提出了基于Fisher准则线性判别分析(LDA)、基于口袋感知机学习算法(PPLA)和以LDA解为初值AUC求解算法等三种简化求解方法。最后通过仿真和实测实验验证了以上方法的有效性和鲁棒性,结果表明以LDA解为初值的AUC求解算法综合性能最佳。

Recently,the diagonal loading detector(DLD),which combines the advantages of the polarimetric whitening filter(PWF)and the polarimetric detection optimization filter(PDOF),is proposed for the detection of weak target in polarimetric synthetic aperture radar(Pol⁃SAR)images in complex sea clutter environment.It breaks the tar⁃get detection performance limitation in non⁃Gaussian clutter background by choosing the appropriate linear weights.How to obtain the optimal weights of this linear combination is the difficulty of this method.In this paper,a linear combi⁃nation based polarimetric diagonal loading detector is reconstructed,and based on this,an optimal mathematical model of the linear combination is given by evaluating the detection performance with the area under curve(AUC),and an optimal weighting algorithm based on this criterion is proposed.Three simplified algorithms are proposed to speed up the solution based on linear discriminant analysis(LDA),pocket perceptron learning algorithm(PPLA),and AUC solving algorithm with the solution of LDA as the initial value,respectively.The effectiveness and robustness of the above methods are verified by simulated data and measured data,and the results show that the AUC solving algorithm with the solution of LDA as the initial value has the optimal performance.