NCIE在多特征选择及SAR目标识别中的应用

针对合成孔径雷达(SAR)图像目标识别问题，采用非线性相关信息熵(NCIE)进行多特征选取进而实现分类。基于混合高斯模型对SAR图像提取的各类特征进行概率建模，采用KL散度评价不同特征之间的相似度。采用非线性相关信息熵评价不同特征组合的相关性，根据最大熵值确定最优特征组合。对于选取的多类特征，基于联合稀疏表示模型进行表征和分类。利用MSTAR数据集对提出方法在标准操作条件和扩展操作条件下进行测试，结果验证了其有效性。

Application of nonlinear correlation information entropy to selection of multiple features and SAR target recognition

The multiple features are selected and classified based on Nonlinear Correlation Information Entropy(NCIE) for the target recognition problem of Synthetic Aperture Radar(SAR) image. The Gaussian mixture model is employed to model the probability distributions of different kinds of features and then the KL(Kullback-Leibler) divergence is utilized to evaluate the similarity among different kinds of features. The NCIE values of different combinations of features are calculated and the one with the maximum entropy is chosen as the optimal. The joint sparse representation model is employed to represent and classify the selected features. Experiments are conducted based on the MSTAR data under the standard operating condition and extended operating condition. The results show the effectiveness of the proposed method.