Interpretation of satellite images using computational intelligence

Abstract

The proposed research work involves interpretation of color satellite images obtained from Landsat and Bhuvan databases. Two datasets are created from these databases consists of 100 images each. The digital information extracted from the Red, Green, Blue and Near Infrared bands of these digital images is used for further processing. Landsat images are downloaded as .jpg or .tiff formats where as Bhuvan images are available in the form of band information in .tiff format. These bands are separately collected and Red, Green 2 and Blue bands are combined to form the database images. These images are selected with three macroclasses containing seven class categories. The macroclasses include land, vegetation and water. The seven class categories include residential land, commercial land, grasslands, evergreen forest, mixed forest, sediments and clear water. The algorithms in proposed research are developed and implemented using programming language in MATLAB (MATrix LABoratory) along with some built-in library functions and Open Source Library for Support Vector Machine, LIBSVM. MATLAB is used for image analysis, graph plotting functions, programming and classification with Neural Networks. LIBSVM is used for carrying out the classification task using Support Vector Machine (SVM). Microsoft Excel and Snipping Tool are used for analysis and pictorial representations. newline