A Road Extraction Approach Based on Fuzzy Logic for High-Resolution Multispectral Data

In this paper, we present an approach based on information fusion for road networks extraction from high-resolution multi-spectral satellite image data that builds upon pixel-based fuzzy logic segmentation approach. This road networks extraction system includes three different modules: geometrical features processing based on local context; segmentation based on fuzzy classification; angular texture signature to refine road clusters. Experimental results show that this method is efficient and distinct in the discrimination of roads and other spectral similar land cover classes.