Multi-Temporal Landsat ETM+Image Fusion Based on ARW for Improved Water Extraction

Rapid and accurate extraction of water is important for investigation and macroscopic monitoring of water resources, wetland protection, coastal line change, and flood inundated area evaluation and flood disaster evaluation. In recent years, more and more researches focus on water extraction using images such as multispectral images and achieved good results. However, when using multi-spectral images, as they are affected by weather (such as heavy rain, snow, and clouds), the precipitation of wet and dry seasons, as well as the growing and withering of vegetation planting in water area, single multi-spectral image can hardly acquire water information as complete as possible even though high level image processing and interpretation technologies are available. Therefore, it is necessary to merge multi data sources with different features together to take use of their complementary information so as to improve water extraction results. A multi-temporal image fusion method based on accuracy-ratio weight is proposed. We analyze the water extraction result of Landsat ETM+ images collected in spring, summer, autumn, and winter respectively, and find out their advantageous and disadvantageous information for water extraction, based on which, a multi-temporal image fusion method is designed, which successfully solves the problem of extraction of plant growing water area and water filling and loss in wet and dry seasons. The proposed multi-temporal image fusion method is proved to have the highest water extraction accuracy in this study.