组合核函数支持向量机高光谱图像融合分类

发布于:2021-10-14 09:53:58

第19卷第4期 2011年4月 光学精密工程 Optics and Precision Engineering Vol-19 No.4 Apr.2011 文章编号 1004—924X(2011)04—0878—06 组合核函数支持向量机高光谱图像融合分类 高恒振,万建伟,粘永健,王力宝,徐湛 (国防科技大学电子科学与工程学院,长沙湖南410073) 摘要:针对高光谱图像分类,提出了一种利用组合核函数融合目标光谱域和空域信息的支持向量机学*算法。该算法首 先用主成分分析方法对高光谱图像进行特征提取和降维,用虚拟维数估计策略预估原始图像的本征维数,并且在预估的 基础上确定要保留的主成份分量数目;然后用数学形态学操作在选取的主分量图像上提取目标的形态信息,得到扩展的 空域形态矢量。最后,通过不同的组合策略,构造组合核函数,从而在分类器中引入空域信息,和原有的谱域信息一起, 利用支持向量机进行分类。高光谱数据实验表明,在训练时间没有显著差别的情况下,总体分类精度和Kappa系数均提 高了2%左右。实验表明,本文提出的方法较单独使用谱域或空域信息进行分类具有一定的优越性。 关 键词:高光谱图像;图像融合;数学形态学;组合核函数;支持向量机 文献标识码:A doi:lO.3788/OPE.20111904.0878 中图分类号:TP751.1 Fusion classification of hyperspectral image by composite kernels support vector machine GAO Heng-zhen,WAN J ian-wei,NIAN Yong-j ian,WANG Li—bao,XU Zhan (School of Electronic Science and Engineering,National University of Deferise Technology,Changsha 410073,China) Abstract:For hyperspectral image classification,a Support Vector Machine(SVM)algorithm with composite kernels was presented to fuse both the spectral information and spatial information of the to extract image.The algorithm adopts Principal Component Analysis(PCA)algorithm feature and reduce the dimension for hyperspectral image,and rithm to uses the image the Virtual Dimension(VD)algo- estimate the Intrinsic Dimension(ID)of the image.Then,the remained number of Principal on Components(PCs)was determined ted by mathematical the basis of the ID.Furthermore,spatial features were extrac— remained PCs,and the Extended morphology from the Morphological Profile to construct (EMP)vector of image was obtained.By combination of different strategies kernels,the spatial information was introduced into the classifier the SVM and based on to composite implement the classification with ex- both the spectral information and spatial information.Hyperspectral image periments indicate that the overall accuracy and Kappa coefficients of the proposed approach increase about 2%without increasing the training time obviously.Compared with the classifiers only using the spatial or spectral information,the proposed method shows a lot advantages. 收稿日期:20lO一10-18;修订日期:2010-11—23. 基金项目:国家自然科学基金资助项目(No.40901216);湖南省研究生科研创新项目(No.CX20108020);国防科技 大学博士研究生创新基金资助项目(No.B100402) 万方数据

相关推荐

最新更新

猜你喜欢