Machine learning approach towards satellite image classification

Classification of the image is an integral part of the digital picture and plays a very significant role in the development of remote sensing technologies. Thus the need to find sophisticated algorithms and methods have shown great interest over the years in solving classification issues. Remote sensing was implemented globally for the local usage of sophisticated satellite networks and sensors. However, the requirement to provide data and decision-taking was a major obstacle. It would be vitally necessary to implement machine learning methods for classification purposes to support the Graphics Processing Unit (GPU) systems to work faster. This paper proposes the supervised technique of machine learning systems, such as K-nearest neighbor (KNN), Artificial Neural Network (ANN), and Decision Tree, in this regard. KNN algorithm is based on matching and averaging non-local neighborhoods. Neural nets are inspired by the learning process that takes place inside human brains while instances are categorized in the decision tree by sorting them down the tree from the root to some leaf nodes. Preliminary findings by performing extensive experiments on satellite image dataset suggest that the proposed classification system may be a competitive alternative in terms of classification and accuracy over current feature-based extraction schemes.