Object-oriented land cover extraction in Changbai Natural Reserve from IKONOS image

Land use/cover is the most prominent landscape sign of the earth surface system. Remote sensing technology provides the land use/cover information quickly and precisely. With the recent high-resolution commercial satellites successfully launched, the image may provide abundant geometrical and spatial information, including texture, structure and shape of ground target and so on. Traditional pixel-based classification approach is greatly limited and it will result in low precision of classification. In this study, taking Changbai Natural Reserve in northeastern China as an example, object-oriented approach was used to extract land cover classes based on eCognition. A model was proposed for the analysis of the most suitable scale of image segmentation. The experiment result shows that object-oriented approach has a great deal of advantage such as high efficiency, high precision, and more impersonality and so on. The precision of object-oriented approach under the most suitable segmentation scale was 87.25%, superior to the traditional pixel-based approach.