Ensemble Subspace Discriminant Classification of Satellite Images

Abstract

Classification is a very important area in satellite remote sensing. If classification process is failed it leads to wrong interpretation of information. To decide whether a classifier is an efficient one or not, it necessary to validate with original information by taking ground truth points of the scene. Support vector machine (SVM), maximum likelihood (ML), K-means, K nearest neighbor (KNN), random forest (RF), etc are present models. The accuracy and other quality parameters obtained with the above said models are not meeting present need. This paper it is proposed a neural network (NN) and ensemble subspace (ES) technique based method to enhance the accuracy and other quality parameters of the classification process. The performance and quality parameters of the ensemble method is compared with state of art classification techniques for low resolution images.