中文摘要：

为避免传统SAR图像特征分类算法中所需的目标方位角精确估计,提出了一种新的基于稀疏表示与空域金字塔环形描述相结合的SAR目标分类方法.该方法引入bag of features思想,利用密集采样SIFT描述特征训练过完备字典,对训练集和测试集同时进行稀疏编码并构造空域金字塔环形描述,得到旋转不变特征,最后输入线性SVM分类器进行分类.MSTAR实测数据的对比实验表明,在无需目标方位角估计的前提下,所提出的算法识别率达到96%以上,取得了很好的目标分类效果.

英文摘要：

Traditional feature based SAR target classification methods require explicit pose angle estimation.To avoid this problem,a SAR target recognition method based on sparse representation and spatial pyramid rings was proposed.The method extended the original bag of features approach to SAR image processing.The dictionary was trained from dense sampled SIFT descriptions.The sparse coding technique and spatial pyramid rings expression were used to gain rotation invariant features.Moving and Stationary Target Acquisition and Recognition public database was used in the classification setup.Results from classifying three categories demonstrate that the performance of the proposed algorithm is superior to the ones using SVM classifier with other dimension reduction techniques,with a classification rate of above 96%.