ISAR高分辨成像算法正则化系数的优化

Optimization of Regularization Coefficient for ISAR High Resolution Imaging Algorithm

以稀疏约束最优化逆合成孔径雷达(Inverse Synthetic Aperture Radar,ISAR)信号模型为基础,基于最大后验概率估计的误差精度和ISAR图像的稀疏性,提出了一种正则化系数优化方法。该方法结合压缩感知理论,基于L1范数建立了稀疏约束的ISAR最优化信号模型。该模型可在任意正信噪比情况下根据每次迭代结果的稀疏度实现正则化系数的自适应调整,且只需少数迭代循环就能确定比较稳定的系数,实现图像的高分辨重建。该方法避免了因无法直接确定最佳系数而重复尝试的复杂过程,提高了算法的实现效率。在求解该最优化模型时,对傅里叶矩阵相乘进行了算法简化并结合共轭梯度下降法降低了算法运算的复杂度。

Based on the sparse constrained optimal inverse synthetic aperture radar(ISAR)signal model,a regularization coefficient optimization method is proposed,based on the error accuracy of maximum posterior probability estimation and the sparsity of ISAR images.Combining the compressed sensing theory,the sparse constrained ISAR optimal signal model is established based on L1 norm.The regularization coefficient can be adaptively adjusted by the model according to the sparsity of each iteration result under arbitrary noise.Only a few iteration cycles are needed to determine the relatively stable coefficient to achieve high-resolution image reconstruction.The complicated process of repeated attempts is avoided because the optimal coefficients cannot be determined directly.The efficiency of the algorithm is improved.The algorithm of Fourier matrix multiplication is simplified and conjugate gradient descent method is used to reduce the complexity of the algorithm.