Assessment of building heights from pléiades satellite imagery for the Nyarugenge sector, Kigali, Rwanda
Quantitative and qualitative information about land-use and buildings are important for spatial planning decision processes. The proposed study determines the suitability of multispectral information and stereoscopic-derived heights of Pléiades tri-stereoscopic satellite imagery for the identification of building footprints and the assessment of building heights. The study area is the Nyarugenge sector in the center of the City of Kigali. A tri-stereoscopic Pléiades satellite scene was used to process a digital surface model. An object-based image analysis provided the basic geometries and respective variables, which served as input features for a Support Vector Machine based classification. The comparison of the building footprints with ground reference information indicates a high accuracy. The assessment of building heights led to mixed results. The heights of buildings were derived with medium to high accuracies for building blocks with a well-developed road network or open spaces, which allowed the assessment of the elevation of the terrain surface. In densely built-up areas, with a lack of road accessibility and located mainly on steep slopes, the derived heights of single buildings or building blocks did not produce satisfying results. However, for urban planners, policy and decision makers, the method yields a high potential in analyzing and monitoring urban areas.
Keywords - building extraction, DSM, remote sensing, SVM, Pléiades