Research on Land Use/Cover Classification Based on GF-1 and Multi-Source Data Combination

Based on the NDVI time-series data, NDWI data, MNDWI data and some other index data which were obtained from the GF-1 multi-temporal data as well as the Landsat8 OLI images and DEM data, the rules of extracting the spatial, multi-temporal and shape features of land objects were derived. A land use/cover classification method of complex terrains based on the GF-1 data was constructed according to those rules and the multi-layer information extraction method. With Guangzhou as the test area, the methods as well as the maximum likelihood, the minimum distance method and the land use/cover classification method of complex terrains based on the GF-1 data were used to class the land used/cover. The results showed that the overall accuracy of classifying the use/cover of land of complex terrains based on the GF-1 data is 85.86%. Besides, the accuracy of extraction of some land objects goes higher than 95%. Compared with the maximum likelihood method and the minimum distance method, this method increases the accuracy by 4.62% and 12.24% respectively, which shows that it can improve the application of GF-1 data in land use/cover classification and increase its accuracy.