基于经验模态分解NCC特征选择的高光谱图像分类算法

Hyperspectral images classification based on NCC feature selection of empirical mode decomposition

采用非线性相关系数(nonlinear correlation coefficient,NCC)进行经验模态分解后的特征选择,通过图像特征(内固模态函数和图像趋势)的非线性相关系数进行特征选择(NCC-Feature Selection),进一步提出了基于非线性相关系数经验模态特征选择的高光谱图像分类算法(2D-EMD-NCC-SVM)。仿真结果证实,算法可选择出可分性强、信息量大的高光谱图像特征,提高高光谱图像的分类精度和分类速度。

In remote sensing classification,empirical mode decomposition improves the performance of support vector machine（SVM） classification.However,there have been few researches on feature selection after empirical mode decomposition.A method of feature selection for empirical mode decomposition based on nonlinear correlation coefficient（NCC） is proposed,and SVM classification based on NCC feature selection of empirical mode decomposition for hyperspectral images is proposed（2D-EMD-NCC-SVM）.Experimental results show that NCC-feature selection can select features with strong separability and rich information,and 2D-EMD-NCC-SVM can significantly increase the classification accuracy and the classification speed.