基于EMP与混合核SVM的高光谱遥感影像分类

Hyperspectral Remote Sensing Image Classification Based on EMP and SVM with Composite Kernel

传统的SVM模型采用同一映射形式的单核模式对叠加的空间特征和光谱特征进行处理,往往无法得到理想的结果,为了解决该问题,提出了一种基于扩展的形态学剖面(EMP)与混合核SVM的高光谱遥感影像分类方法。该方法首先通过EMP有效提取空间信息,再采用不同的核函数处理空间信息与光谱信息,最终完成混合核SVM的高光谱影像分类。对多种组合形式的单核以及多核SVM模型进行了对比分析,结果表明,该方法具有较高的适应性,对于高光谱遥感影像的分类精度较高。

The traditional SVM model directly uses the single kernel mode of the same form to process the superimposed spatial features and spectral features,it is often unable to obtain satisfied results.In order to solve this problem,we proposed a hyperspectral remote sensing image classification method based on extended morphological profile(EMP)and SVM with composite kernel.In this method,we extracted the spatial information effectively by EMP at first.After that,we used the different kernel functions to process spatial information and spectral information.Finally,we completed the hyperspectral image classification of SVM with composite kernel.We compared and analyzed the single kernel and composite kernel SVM models with multiple combination forms.The result shows that the proposed method can achieve higher classification accuracy compared with the traditional SVM model.