TIN surface reconstruction from multiple calibrated aerial images in urban areas

This paper describes a framework for surface reconstruction in urban areas from digital multiple aerial calibrated images. A dense and reliable DSM (digital surface model) is first generated and used to split the scene in two kinds of areas of interest - ground and buildings. The surfaces within these areas are modeled under TIN (triangular irregular network) form by a constrained Delaunay triangulation of a selection of 3D points and line segments provided by reliable true multi-image feature extractions. The DSM are reliable, accurate, and correctly reproduce landscape morphology (discontinuities, steep slopes, slope breaks, microstructures, etc.).