Fusing multi-source urban maps under consideration of spatial resolution and environmental characteristics

摘要

In this work a decision level fusion algorithm incorporating multispectral- and SAR- based urban maps from different spatial resolutions is presented. A variety of urban maps from SAR as well optical EO data with spatial resolutions ranging between 12 and 75 meters are combined in order to produce accurate urban area extractions of the Chinese 'Megacities' of Beijing and Guangzhou. The influence of spatial posting of the data, climate region and season at image acquisition has been evaluated and reflected in optimized weights of the input data in the fusion process. Therefore, the weights are determined based on the assumption that the usability and reliability of EO data products depends on the intended scale of analysis as well as on the fact that their sensor-specific advantages and limitations are strongly connected to the environmental characteristics of the area under investigation. Results show that improved built-up area extractions can be achieved with the proposed method.