Abstract

We used Advanced Visible Infrared Imaging Spectrometer-Next Generation (AVIRIS-NG) hyperspectral data for deriving automated layered igneous intrusion maps by implementing the support vector machine (SVM) algorithm. We proposed a spectral analysis based approach to identify a set of optimum input spectral bands for deriving SVM-based maps of layered rocks of Sittampundi Layered Complex, India. We used three SVM models: (a) in the first model, we implemented SVM using nine spectral bands for deriving spectral indices to delineate different rocks; (b) in the second model, we used three Principal Component (PC) bands, which suitably preserved the spectral variance of all the bands used for deriving spectral indices of rocks; and (c) In the third model, all the spectral bands of AVIRIS-NG were used as input for the SVM model. We found PC based SVM model was superior as compared to the other two models in deriving automated map of layered rocks.

摘要

我们使用高级可见红外成像光谱仪 - 下一代 (AVIRIS-NG) 高光谱数据，通过实施支持向量机 (SVM) 算法来推导自动分层火成侵入图。我们提出了一种基于光谱分析的方法来确定一组最佳输入光谱带，用于导出印度 Sittampundi 层状复合体的基于 SVM 的层状岩石图。我们使用了三个 SVM 模型：(a) 在第一个模型中，我们使用 9 个光谱带实现了 SVM，用于推导光谱指数以描绘不同的岩石；(b) 在第二个模型中，我们使用了三个主成分 (PC) 波段，它适当地保留了用于推导岩石光谱指数的所有波段的光谱方差；(c) 在第三个模型中，AVIRIS-NG 的所有光谱带都用作 SVM 模型的输入。