HY1C/D-CZI Noctiluca scintillans Bloom Recognition Network Based on Hybrid Convolution and Self-Attention

Accurate Noctiluca scintillans bloom (NSB) recognition from space is of great significance for marine ecological monitoring and underwater target detection. However, most existing NSB models require expert visual interpretation or manual adjustment model thresholds, which limits application in operational monitoring. To address these problems, we developed a Bloom Recognition Network (NSBRNet) incorporating an Inception Conv Block (ICB) Swin Attention (SAB) based on the latest deep learning technology, where ICB uses convolution to extract channel local detail features, SAB self-attention global spatial features. The was applied Coastal Zone Imager (CZI) data onboard Chinese ocean color satellites (HY1C/D). results show that NSBRNet can automatically identify using CZI data. Compared with other common semantic segmentation models, showed better performance precision 92.22%, recall 88.20%, F1-score 90.10%, IOU 82.18%.