GF-1/6 Satellite Pixel-by-Pixel Quality Tagging Algorithm

The Landsat and Sentinel series satellites contain their own quality tagging data products, marking the source image pixel by with several specific semantic categories. These products generally categories such as cloud, cloud shadow, land, water body, snow. Due to lack of mid-wave thermal infrared bands, accuracy traditional detection algorithm is unstable when facing Chinese Gaofen-1/6 (GF-1/6) data. Moreover, it challenging distinguish clouds from In order produce GF-1/6 satellite pixel-by-pixel this paper builds a training sample set more than 100,000 pairs, primarily using Sentinel-2 Then, we adopt Swin Transformer model self-attention mechanism for tagging. Experiments show that model’s overall reaches level Fmask v4.6 10,000 samples, can between snow correctly. Our meet requirements “Analysis Ready Data (ARD) Technology Research Domestic Satellite” project.