An urban open space extraction method: Combining spectral and geometric characteristics

This paper introduces an advanced method based on remote sensing and Geographic Information System for urban open space extraction combining spectral and geometric characteristics. From both semantic and remote sensing perspectives, a hybrid hierarchy structure and class organization of open space are issues and mapped from one to another. Based on per-pixel and segmentation mechanism separately, two classification approaches are performed. Owing to prior of spatial aggregation and spectral contribution, the segmentation-based classification exhibits its superiority over a pixel-based classification. Finally a GIS-based post procedure is hired to eliminate some unsuitable open space components in both spatial and numerical constraints on the one hand, and separate open space some fabrics from fused remote sensing classes by defining their Shape Index on the other hand. The case study of Beer Sheva based on ASTER data proves this method is a feasible way for open space extraction.