Surface Feature Extraction on the Basis of Object-oriented Remote Sensing Classification Methods in Manas River Basin

Abstract

Remote-sensing (RS) images were extracted by using object-oriented remote sensing classification methods. This study combines RS and Geographic Information System to conduct multilevel segmentation and classify of the remote sensing image of Manas watershed. The e-Cognition system was selected to define the knowledge base following the classification system. The results show that the overall nicety of grading can reach 97.37% and that the Kappa coefficient is 0.9706. These results show that land use can be described and extracted by using high spatial resolution remote-sensing images.