基于形态学自适应分块的高分辨SAR多特征增强算法

传统基于l1范数正则化算子(least absolute shrinkage and selection operator,LASSO)模型的合成孔径雷达(synthetic aperture radar,SAR)压缩感知类稀疏成像算法易丢失弱散射点.基于扩展型组LASSO系列模型的算法虽可增强SAR结构特征以保留弱散...

Multi-feature enhancement algorithm for high resolution SAR based on morphological auto-blocking

The conventional compressive sensing sparse imaging algorithm of synthetic aperture radar(SAR)based on the least absolute shrinkage and selection operator(LASSO)is easy to lose weak scatters.Although the imaging algorithm based on extended group LASSO is capable of strengthening weak scatters of structure of target of interests,it is inflexible to adaptively block the groups for enhancing the target feature,since only the Euclidean distance is adopted for the blocking operation,and the quality of the SAR imagery is poor.In this paper,a morphological auto-blocking alternating direction method of multipliers(MAB-ADMM)algorithm is proposed for the structural feature enhancement of high-resolution SAR imagery.A mixture of ■M/■F norm and an ■1 norm are introduced to represent the structural and sparse features,respectively,and a coordination process is carried out to enhance the features cooperatively.Because the proposed MAB-ADMM algorithm employs the morphological blocking scheme based on the geodesic distance,it is capable of capturing the contour of target of interests effectively,and the accuracy and integrity of the enhancement can be improved.In the experiment,both the simulated and raw SAR data are used for the verification of the proposed algorithm.Comparisons with the conventional algorithm are performed to show the superiority of the proposed algorithm.The phase transition analysis is carried out to evaluate the performance of the proposed algorithm quantitatively.Also,the MSTAR data set is employed to verify the performance of the proposed algorithm in preparing for the SAR target classification.