Feature Selection and Comparative Analysis of the Supervised Learning Model for Hyperspectral Image Classification

In remote sensing image classification, really it is an intimidating when kernel supervised learning approaches stands in need of adequate amount of training samples. Often there is a vital problem for definition and acquisition of reference data. For Hyperspectral image classification, improved spectral information is required to make it suitable for ground object identification. In this paper, Support Vector Machine with RBF kernel (KSVM) and the spectral angle mapper (SAM) are used for performance comparison in terms of classification accuracy in Hyperspectral image classification. Kernel support vector machine is more preferable for the mastery to generalize better hyperplane when limited availability of training samples and separate the classes competently in a new dimension feature space. Experiments are performed on NASA Airborne Visible Infrared Spectrometer (AVIRIS) image and it shows KSVM outperforms SAM and obtains the highest accuracy. Due to more well-conditioned against the outliers, KSVM significantly reduced the classification complexities than SAM.