Land Cover Extraction Based on Object Oriented Analysis

Abstract

Conventional pixel-based image analysis is used frequently to extract land cover feature from remotely sensed data. Therefore only the spectral information is used in the process of classification. However ,object-oriented image analysis can overcome the limitation stated above. It is based on fuzzy logic functions, uses multiscale image segmentation techniques, allows the integration of a broad spectrum of different object features,such as sepctral values,shape and texture. In the paper, the need of land cover classification is satisfied by the result of object oriented analysis in the middle of Fenjie county of the Three Gorges Area, as well as the feature extraction. At the mean time , there is a validation by using in-situ observation ,which shows that the overall accuracy is 89.2%.