Comparison of Supervised Classification Techniques for High-Resolution Optical Aerial Image

Image classification is considered as a very effective tool to monitor the several types of objects or classes present in the area of investigation. Classification of the images can be performed in two manners, i.e. supervised classification and unsupervised classification. Here we have investigated comparison of supervised classification techniques through a post-classification confusion matrix which evaluates our classification performance in term of accuracy assessment and kappa value. In this research work have performed the classification of the high resolution optical aerial image followed by accuracy assessment and development of a relationship through the accuracy obtained from the individual case of classification. Here we have compared four supervised classification techniques which include Parallelepiped classification, Minimum distance classification, Mahalanobis distance classification and Maximum likelihood classification scheme. We have obtained the accuracy of classification followed by estimation of the kappa value for each classification. Finally, we have developed a relationship between the kappa value and the overall accuracy.