基于监督非相干字典学习的极化SAR图像舰船目标检测

提出了一种结构化非相干字典学习算法 (Structured incoherent dictionary learning, SIDL),并将该方法应用于极化SAR (Polarimetric synthetic aperture radar, PoLSAR)图像舰船目标检测. 在字典学习阶段,构建了一个新的目标函数,为了降低子字典对交叉样本的稀疏表示能力, 将子字典对交叉样本的重构能量约束及子字典互相干性约束加入到字典学习目标函数中. 通过这两个约束, 降低了子字典对交叉样本的表示能力,目标和杂波的极化特征矢量在学习获得的字典下具有良好的区分特性. 该方法不依赖于目标后向散射能量,只利用学习获得的极化字典,根据测试样本在极化字典下的稀疏表示进行目标的检测. 实验采用RADARSAT-2数据进行了验证,对比实验结果表明,本文提出的方法可以更好地抑制杂波,对弱小目标实现检测,获得了更好的检测效果.

Supervised Incoherent Dictionary Learning for Ship Detection withPolSAR Images

This paper presents a supervised structured incoherent dictionary learning (SIDL) algorithm and applies it to ship detection using polarimetric synthetic aperture radar (PolSAR) images. In the dictionary learning stage, to weaken the sub-dictionary reconstruction ability to cross-samples, a penalty of cross-representation and a penalty of sub-dictionary cross-incoherence are added to the objection function. So as to restrict the reconstruction of target samples by the clutter sub-dictionary. Then, several statistics are defined with the sparse representation under the learned supervised structured incoherent dictionary. The proposed detection algorithm of the proposed detection algorithm does not depend on the intensity but on the quality of the learned dictionary. The performance is verified using the RADARSAT-2 data set, by comparing it with the partial target detection (PTD) method, the reflection symmetry detection method (RSD), the polarimetric whitening filter (PWF) and the intenstity detector (SPAN). The experimental results show that the proposed method can suppress the clutter better and achieve a better detection performance.