关键词: ALOS, 土地覆盖, 分类, 景观格局

Abstract: Land cover classification is processed by Mahalanobis Distance method, maximum likelihood method, and support vector machine method with confusion matrix assessing the classification accuracy. The results show that the maximum likelihood method and support vector machine method are effective. Maximum likelihood method is taken as an example for land cover classification performance analysis by introducing the Normalized Difference Vegetation Index（NDVI） and GLCM texture features. Results reveal that the classification accuracy is improved to varying degrees when NDVI, CONTRAST, MEAN are incorporated in classification process. Further combination of the three features with the original band represents the highest classification accuracy. The general landscape pattern characteristics are quantified based on  classification results and landscape metrics. The results show that cropland and human settlements are dominating the landscape with an area ratio of 82%, which implies high human pressure and ecological risk in the landscape. It is, therefore, necessary to strengthen land use planning and management in the desert-oasis transitional area in the middle reaches of the Heihe River. Specifically, concrete measures are urgently needed to slow down and regulate cropland and residential area expansion with improved land use efficiency. At the same time, ecological conservation and rehabilitation need to be enhanced to increase landscape resilience to disturbances.

Key words: ALOS, land cover, classification, landscape pattern