High-Accuracy Mangrove Mapping Using Multi-Feature Fusion with Sentinel-2 Data: A Case Study of Yingluo Bay

Mangrove is an evergreen woody plant community, which plays an important role in maintaining good coastal hyrdology , improving ecology and creating carbon sink. The purpose of this study is to explore an efficient and accurate method for analyzing mangrove vegetation information to support the proper understanding of the changes that occur in mangrove cover. In order to achieve this goal, this research uses the Yingluo Bay Sentinel-2 data, and uses the sen2res model to super-resolution reconstruct the preprocessed data to 10 meters. The image data is reconstructed by constructing the index between the visible light band and the red edge band, as well as the band combination with other commonly used vegetation indexes. Then, through correlation analysis, the low correlation and negative correlation index bands are retained based on NDVI to reduce redundancy and improve diversity. Finally, the contribution of texture features in the classification process is evaluated by experiments. The experimental results show that this method can extract mangrove efficiently and accurately and calculate its area. The kappa coefficient of land feature classification is 97.6% and the relative accuracy of mangrove area is 95.1%. In the ranking of contribution degree, texture features contribute the most in this random forest classification, and the contributions of spectral features and index features are also more prominent. Experiments show that the method discussed in this paper can improve the classification accuracy, and the index feature and texture feature should be added in the land cover classification . This Research has substantial significance for scientific support of mangrove resources statistics, and provides reliable methods and data support for mangrove ecological monitoring and detection.