Application of the MLP-FNN model in the classification of coral reef surface cover: a case study in the South China Sea

Using multi-source satellite remote sensing technology, this study extracted the instantaneous geospatial morphology of existing coral reefs in the South China Sea. A multi-layer feedforward neural network (MLP-FNN) model was then employed to conduct an in-depth quantitative analysis of the spatial distribution, trend changes in characteristic parameters, and land cover classification of the coral reefs in the Spratly Islands. The results indicate that high-resolution remote sensing data from Sentinel-2 satellites can effectively identify the spatial characteristics and land cover types of coral reefs and islands far from the mainland. The study found that between 2019 and 2022, the area of coral reef clusters in the study region decreased significantly, while the area of coral sedimentation zones continued to increase. Additionally, there were significant spatial distribution differences in coral reef clusters across different orientations of the atolls.