SAM-Classification-of-satellite-images

The project aims at explaining the usage of SAM algorithm for satellite image classification. Hyperspectral Image provides pixel spectrum that fetches detailed information about a surface to identify and distinguish between spectrally similar (but unique) materials. The Hyperspectral Image sensor placed on board the Remote Sensing Satellite captures Hyperspectral Images with various bands of spectrum. Experiments are carried out for the implementation of Spectral Angle Mapper (SAM) on Hyper- spectral Images for classification of pixels on the surface. The false color composite of the image is also obtained for better visualization of surface differences. The Hyperspectral Images of various bands are stacked one after the other to form three-dimensional Cube of images for SAM implementation. SAM is a supervised classification algorithm which identifies the various classes in the image based on the calculation of the spectral angle. The spectral angle is calculated between the test vector built for each pixel and the reference vector built for each reference classes selected by the user. Results are obtained to read and reorganize multiple 2-D datasets into a single compact 3D dataset cube. The reference vector is built for performing SAM classification and the angle between the reference vector and pixel vector is calculated to compare with the determined threshold angle value. The color coding is then applied to distinguish between the various classes that have been recognized by the SAM algorithm. Hence using SAM, Hyperspectral images are analyzed to extract thematic information such as land-cover, water bodies, and clouds.