DETECTION OF RIVER WATER SURFACE CHANGES IN NANCHANG BASED ON GRAPH GENERATIVE STRUCTURE-AWARE TRANSFORMERS

River surface change detection is a vital technology for watershed monitoring, enabling real-time identification of dynamic hydrological variations through remote sensing image analysis. However, accurately identifying river surfaces remains a challenge, as it requires simultaneously considering both local and global information within the river area. Recently, we developed a graph generative structure-aware Transformer (GraphGST) for hyperspectral image classification. Considering that the model can efficiently capture both local and global contextual dependencies without the need for additional operations, this paper extends the model to achieve river surface change detection for hyperspectral images.These results demonstrates the superiority of our approach in refining water body contour recognition and enhancing overall change detection performance.