Three-dimensional singular spectrum analysis for precise land cover classification from UAV-borne hyperspectral benchmark datasets

• DOI: 10.1016/j.isprsjprs.2023.07.013
• Publication Date: 2023-08-03T10:43:22Z
• Authors: Hang Fu; Genyun Sun; Li Zhang; Aizhu Zhang; Jinchang Ren; Xiuping Jia; Feng Li
• Keywords: Land Cover; Benchmark (surveying); Disjoint sets

Qingdao UAV-borne HSI (QUH) dataset consists of three sub-datasets: QUH-Tangdaowan, QUH-Qingyun, and QUH-Pingan, which are freely available as benchmarks for precise land cover classification. The QUH-Tangdaowan dataset: it contains a variety of irregular land covers with very similar spectral curves, such as four vegetation species: Coniferous pine, Buxus sinica, Populus, Ulmus pumila L, and three pavements: Flagging, Boardwalk, Gravel road, which undoubtedly poses a great challenge for precise classification. The QUH-Qingyun dataset: it has areas obscured by building shadows, including classes such as Trees, Car, and Asphalt road. The interference of shadows can significantly reduce the accuracy of identification and can be used to assess the robustness of the classification methods. The QUH-Pingan dataset: it has a relatively regular distribution of land cover, but there are large differences in size: Seawater, Road at large scales, Trees, Floating pier, Brick houses, Steel houses at medium scales, and Ship, Car at small scales, which are the most difficult to identify. This dataset can be used to assess the ability of classification methods to portray targets at different scales.