一种基于NPA的加权“1 V m”SVM高光谱影像分类算法
An Algorithm of Weighted “1 V m” SVM Multi-classification for Hyperspectral Remote Sensing Image with NPA
-
摘要: 根据支持向量机(SVM)的计算理论,结合高光谱影像的数据特点,利用最近点算法(NPA)求两类最优超平面,为每类设立一个合理的权指标,提出了基于NPA的加权“1 V m”SVM算法来实现高光谱遥感影像多分类,降低了计算的复杂度和计算量,提高了SVM高光谱遥感影像分类的可操作性和分类效率。Abstract: According to the SVM computation theory and the features of hyperspectral remote sensing(RS) image data,the optimal hyperplane between two classes is computed by the nearest points algorithm(NPA).Reasonable weight indicators are designed for each class and a new weighted "1 V m" SVM based on NPA is proposed to achieve Hyperspectral RS image classification.The new algorithm can reduce the computational complexity and calculation of SVM,and improve SVM feasibilities and efficiencies for hyperspectral RS image classification.Finally,a test was carried out on OMIS image and good results are obtained.
下载:
