Application Of Landslide Information Extraction From High Resolution Remote Sensing Based On Object Oriented Classification Method

Landslide identification based on remote sensing technology has been studied for several decades, but there is still a serious drawback of accuracy. In recent years, with the development of satellite technology, the resolution of satellite images have been largely enhanced while the pixel-based image analysis methods cannot meet the demand of research. But the object-oriented classification methods improves the accuracy and efficiency in landslide identification. Related researches mainly focus on selection of satellite images and methods in identification of landslide.In this paper, the research was conducted on the fused images of Aster and Geoeye using object-oriented classification methods.The main contents and achievements of the study are as following:1. The application of the object-oriented classification methods in landslide indentification and the exploration of the optimal strategies in image segmentation, selection of objects’attributive characters, and classification.2. Gonglingping catchment was selected as the study area, which located in Wudu county, southern Gansu. The bands of ASTER image were analyzed to obtain the optimum composition of spectral bands, then fused with GEOEYE image, which increased the resolution of image.3. The fused images had been conducted to the object-origented image classification and then the better image segmentation was obtained.4. The fused images had been conducted to k-nearest neighbourhood classification algorithm and the classification samples and testing samples were selected to test the precision of classified result.5. Selecting the appropriate attribute to study the classification using rule-based method and evaluate the accuracy of classification.6. At last, the two classification methods were compared, and the classification method for better precision had been selected to extract the landslide disaster information.