Classification of multispectral satellite images using ensemble techniques of bagging, boosting and adaboost

Various methods exist for classification of multispectral satellite images. Very few techniques have tried using ensemble of classifiers using artificial neural networks to increase the accuracy of classification. In this paper the performances of single classifiers using various neural networks classifier is compared with ensemble classifier. Individual neural network used are backpropagation and radial basis function. Classification for same image using ensemble of backpropagation neural networks with change in number of neurons is used. Ensemble is achieved using bagging, boosting and adaboosting techniques. It is observed that the performance of ensemble classifiers is better than individual classifiers. The input image is divided into 8X8 blocks and features used for to train the ensemble network are mean, variance, standard deviation and texture of each block The performance is measured using various parameters such as producer´s accuracy, user´s accuracy, overall accuracy, kappa Coefficient and confusion matrix for different classifiers.