Supervised/Unsupervised Classification of Land cover using Remote Sensed Data

Abstract

Remotely sensed images may be used to predict a wide range of properties at the surface of the Earth including both categorical and continuous properties. Batticaloa District which encompasses different land cover categories including water, barren land, agriculture and forest. This study achieves maximum likelihood supervised classification, Iso clustering unsupervised classification to an Sentinal 2 Data image, and compares the results of these three methods in Batticaloa District. Secondary data was utilized to this study and supervised classification and unsupervised classification methods were engaged in this study. The results of all classifications revealed that agriculture classes were the largest land cover classes in this area. The extent of water bodies was happened approximately equal in all classification process in this selected area.  Barren land cover classes a major difference in all classification and forest area show major deviation in supervised and unsupervised classification. The water body’s classes of resulting supervised classified image are generally acceptable even some water bodies came within the barren land categories. So, training sample polygon should be selected homogenous area in this process because of highly depending the accuracy of supervised classification by sample creator.