2026-10-14T23:42:59Z https://www.scidb.cn/oai

10.57760/sciencedb.30189 2025-12-05T09:40:55Z

2025-12-05 Dragon fruit tree identification and small-sample semantic segmentation dataset based on ultra-low altitude remote sensing doi:10.57760/sciencedb.30189 en Using the Red Dragon Fruit Planting Base in Longping Town, Luodian County, Guizhou Province as the experimental area, a fine semantic segmentation dataset for dragon fruit trees was constructed based on ArcGIS software and manual visual interpretation of sample annotation using unmanned aerial vehicle field aerial data from September 2020. This data set is 4.80G, mainly consisting of 12 unmanned aerial vehicle (UAV) ultra-low altitude remote sensing images with a resolution of 5000 × 5000 pixels, which clearly present the growth environment and morphology of the dragon fruit tree; Manually annotated dragon fruit tree images, shp data, and schematic diagrams provide annotations for model training; Simultaneously covering initial (600 samples), supplementary (2400 samples), and expanded (21593 samples) 224 × 224 pixel drone remote sensing image samples and manually labeled images, as well as 12 identical. shp data for validation. The experiment showed that the U-Net model trained on this dataset achieved high accuracy (98.49% accuracy, Kappa coefficient 0.91) in the recognition task of dragon fruit trees.  ultra-low altitude remote sensing; dragon fruit; small-sample learning; semantic segmentation; precision agriculture Shu Guoxiang Lan Yan Zhang Xiulian Ma Mingyue Li Qianxia Huang Denghong yan li hui PUBLIC https://creativecommons.org/licenses/by-nc-nd/4.0/ dataset http://www.doi.org/10.11922/11-6035.csd.2025.0169.zh Science Data Bank