Abstract: An extended one-class support vector machine (OCSVM) was applied to lithologic mapping from the EO-1 Hyperion hyperspectral data, and it has been evaluated in terms of classification accuracy. First OCSVM was separately used to extract each lithologic unit of interest. The pixel which was classified to different classes simultaneously was then assigned as the class with smallest distance to the hyperplane. In this way, the extended OCSVM can be used for extracting several lithologic units of interest. The extended OCSVM method was used in lithologic classification from the EO-1 Hyperion hyperspectral data in Junggar area, Xinjiang and compared with the spectral angle mapper (SAM) method. The results showed that the extended OCSVM method outperformed the SAM method in lithologic classification. The extended OCSVM is a useful and effective method for lithologic classification from hyperspectral remote sensing data.

Key words: hyperspectral data, OCSVM, SAM, lithologic classification

摘要： 将扩展的单类支持向量机方法运用到高光谱岩性识别中, 并分析和评价该方法的性能。利用单类支持向量机分别提取各个感兴趣的岩性类别, 对于被识别为多个岩性类别的像元, 根据该像元与每个单类支持向量机所确定的分类超平面的距离来确定属于哪一类别, 这样, 利用扩展的单类支持向量机来可提取多个感兴趣的岩性类别。将该方法运用到新疆准噶尔地区的EO-1 Hyperion高光谱数据岩性分类中, 并与传统的光谱角制图方法进行比较。结果表明, 扩展的单类支持向量机方法的岩性分类精度显著高于光谱角制图方法, 是一种新的可用于高光谱数据的岩性分类方法。

关键词: 高光谱, 单类支持向量机, 光谱角制图, 岩性分类