A study on the land cover classification of arid region based on multi-temporal TM images

作者：Li Haoxu^*; Xie Yaowen; Yu Lin; Wang Lijun

来源：2011 3rd International Conference on Environmental Science and Information Application Technology, ESIAT 2011, 2011-08-20 to 2011-08-21.

DOI：10.1016/j.proenv.2011.09.374

• Albedo
• Arid lands
• Bare lands
• Cover types
• Decision tree classifiers
• Ecological environments
• Field investigation
• Gansu province
• High resolution image
• Land cover classification
• Land-cover types
• MNDWI
• Multi-temporal
• Region-based
• Study case
• Temporal images
• TM image
• Vegetation cover
• Vegetation coverage
• Wetness

For the fragile-ecosystem arid land, the macro and integral information about the land cover types and area is important for the restoration and the improvement of the ecological environment. In this paper, the arid Gaotai County of Gansu Province, China, was taken as a study case to discuss classified method of the arid land based on various indexes. Taking the advantages of different temporal images into account, two TM images acquired in October and June respectively were selected to conduct the research. Firstly, a series of indexes such as NDVI, vegetation coverage, MNDWI, Albedo, Wetness and SI were calculated. Secondly, combining Robert algorithms and artificial experience methods, the integrated index threshold for distinguishing vegetation from other land types was calculated. Thirdly, using the threshold and decision tree classifier, the middle-high and low coverage vegetation were extracted from both two images. Fourthly, the vegetation areas from two images were combined to form the total vegetation cover types (Oasis), which was used to mask October images, and the outside of mask area was continually classified into the types such as the salinization land, water, bare land and wetlands and other land cover types with decision tree classifier. At last, combining all cover types together, the land cover classification map of whole county was produced. Compared with the land cover types interpreted from the high resolution images and field investigation points, this method was found achieving the accuracy of 91.5%.

• 出版日期2011