Mangrove Extraction in Shenzhen Area Based on Multi-Source Temporal Remote Sensing Data

In this study, Sentinel-1 image data with all-weather observation characteristics and Sentinel-2 image data with medium and high spatial and temporal resolution were used to explore the optimal feature scheme suitable for coastal wetland classification. The optimal feature combination scheme will be put into the UNet network and applied to the mangrove classification in the coastal area of Shenzhen. In addition, we added DCN structure to the model, which can better extract mangrove features compared with traditional methods. The experimental results showed that the mangrove area in Shenzhen showed an overall increasing trend in the past four years, which proved that the local mangrove restoration process had a certain effect.