Comparison of Supervised Classification Methods On Remote Sensed Satellite Data: An Application In Chennai, South India

This paper presents classification of various land cover types from the raw satellite image using supervised classifiers and performances of the classifiers are analyzed. Geo coded and Geo-referenced remote sensed images from Survey of India, Government of India Topographical maps are used. Prior to classification, Training process to assemble a set of statistics describing spectral response pattern of each land cover type is done. The quality of training plays a crucial role in success of classification. Classification is executed based on the spectral features using Minimum distance to mean classifier, Maximum likelihood classifier and Mahalanobis classifier. Efficiency of Classification results are assessed by using accuracy assessment and Confusion matrix. Performance of Maximum likelihood classifier is found to be better than other two. ERDAS IMAGINE 9.2, the world’s leading geospatial data authoring software