基于主元分析和稀疏表示的SAR图像目标识别

Target recognition of SAR images using principal component analysis and sparse representation

现有的合成孔径雷达图像目标识别方法通常包括图像预处理、特征提取和识别算法3部分。但是,预处理算法的自适应性很难得到保证。提出了一种基于主元分析和稀疏表示的目标识别算法。首先,阐述了稀疏表示和重构的基本理论;其次,提出了基于主元分析和稀疏表示的合成孔径雷达图像目标识别算法;最后,选取MSTAR数据库中的5类合成孔径雷达目标图像进行仿真。结果表明,在没有预处理的情况下,该算法仍能有效地识别目标,与主元分析和三阶近邻的识别算法相比,具有较高的识别率和鲁棒性。

With the existing target recognition algorithms of synthetic aperture radar （SAR） images, image preprocessing, feature extraction and recognition algorithm are usually carried out. The adaptability of the pre- processing algorithm is difficult to be guaranteed. A target recognition algorithm using principal component analysis （PCA） and sparse representation is proposed. Firstly, the basic theory of sparse representation and re construction is presented. Secondly, an SAR image target recognition algorithm is presented using PCA and sparse representation. Finally, an experiment with five kinds of SAR target images in the MSTAR database is given. The simulation results show that this algorithm can still recognize the target effectively without prepro- cessing. Compared with the PCA and the third-order nearest neighbor algorithm, the proposed algorithm has a higher recognition rate and robustness.