Methods for automatic identification and extraction of terraces from high spatial resolution satellite data (China-GF-1)

Terraces are effective measures to control soil and water loss on hillslopes and play significant roles in water storage, soil conservation and increasing of crop yield. With the launch of comprehensive management engineering of transforming slopes into terraces in China, the area of terraces is continuously expanding. To plan for soil and water conservation in a scientific way and ensure rational soil and water conservation measures, it is important to determine a timely and accurate inventory of the spatial distribution of the existing terraces. High-resolution satellite data (China GF-1) was used in this study to automatically identify and extract terraces. The edge characteristics statistics algorithm, template-matching algorithm and the Fourier transformation algorithm were used. The results suggest that the template-matching algorithm, based on high gray features of terraces, has higher identification accuracy and lower extraction miss rate, and is able to determines the complete map spots, which works well for identification and extraction of typical terraces. The Fourier transformation algorithm, based on the texture of the terraces, is easily affected by gray changes on the terrace field surfaces and field ridges and is hardly reliable as a stand-alone technique for terrace identification. In this study, we focused on an area near Hengshan County, Yulin of Shanxi Province, and performed field validation of the terrace automatic identification and extraction algorithms. The results suggest that the accuracy of automatic identification and extraction of the terraces is greater than 80%. The method proposed is helpful to meet the requirements for planning and monitoring soil and water conservation projects.