A Study on Object-Oriented Classification of Land Use/Land Cover of Remote Sensing Image

Land use/land cover classification in the south of the Gaoligong Mountain based on Landsat TM by use of object-oriented classification technique are presented. Through several key steps, such as image pre-processing, feature extract, multi-segmentation, and image classification, Land use/Land cover in the mountainous area is successfully mapped. The accuracy assessment indicates that the object-oriented classification is better than traditional pixel-based method: the overall accuracy and the Kappa coefficient is 88.6% and 0.82 respectively, increasing by 14.9% and 0.22 respectively compared with the pixel-based classification; the object-oriented method can meet the need of land use classification, and can be used as a new technique in land resources survey.