Adaptive Environment Geographic Optimal Transport Based on Remote Sensing Feature Analysis

The fusion of remote sensing data and social media data effectively enhances the spatiotemporal resolution of available datasets, providing assistance in addressing specific issues. For the integration of heterogeneous data, domain adaptation and Geographic Optimal Transport (GOT) methods have offered significant support. However, within the process of data fusion, the climatic environment of the data collection areas is often less considered, resulting in the omission of certain climate characteristics under remote sensing imagery, which typically significantly impact algorithm precision. In this paper, building upon the GOT, we further explore the impact of variations in different remote sensing parameters on data fusion accuracy under specific climatic conditions, leading to modifications in the GOT formula. In experiments conducted for flood scenarios, we have enhanced the accuracy of heterogeneous data fusion in this context by emphasizing the MNDWI term and removing the NDVI term from the GOT equation.