Abstract: Desertification of land was the significant environmental issue threatening human existence. Desertification land area was 1 731 100 km² in 2009, occupying 18.03% of China's total land area. Remote sensing is an effective method monitoring the desertification land. With the development of remote sensing technology, high spatial resolution remote sensing image gradually becomes main data source of resolving the desertification land information extraction. The traditional remote sensing image classification method based on pixel has some difficulties in treating the high spatial resolution remote sensing images, however, the object-oriented classification method might overcame the limit of using the pixel as basic treatment unit. Minqin County, a typical desertification land distribution region, was used as studied object in this paper, the high spatial resolution remote sensing images SPOT5 and 30 m DEM were used as date sources, the multi-scale and multi-level segmentation of remote sensing images was conducted by using the object-oriented classification method. Based on these results, the remote sensing images were subjected to fuzzy image classification based on knowledge by constructing multiset, and thus relatively accurately realize the recognition and extraction of desertification land information. This paper would provide an effective path for resolving the accurate extraction of desertification land information based on the high spatial resolution remote sensing images and also provide scientific proofs for sand prevention and control.

Key words: SPOT5 image, object-oriented classification method, desertification land, information extraction, remote sensing