One-Eye Stereo System for the Acquisition of Complex 3D Building Descriptions

An easy usable system for the semi automatic acquisition of detailed D building descriptions from a multitude of images is provided This approach tackles robustly and e ciently most of the problems of D building reconstruction namely occlusions inverse mapping and noise The D modeling is based on Con structive Solid Geometry CSG and various automated and supporting tools Our experiences on the acquisition of an extensive scene are evaluated Zusammenfassung Monokulares Stereo System f ur die Erfassung kom plexer D Geb audebeschreibungen F ur die semiautomatische Erfassung de taillierter D Geb audebeschreibungen aus einer Menge von Bildern wird ein einfach handhabbares System vorgestellt Auf robuste und e ziente Weise werden die meis ten Probleme bei der D Geb auderekonstruktion Verdeckungen inverse Abbildung und Rauschen angegangen Die D Modellierung basiert auf der Constructive Solid Geometry CSG und zahlreichen automatisierten und unterst utzenden Werkzeugen Unsere Erfahrungen mit einer gro achigen Erfassung werden evaluiert Introduction and Motivation An increased demand of D building descriptions for environmental and city planning air distribution and air pollution simulation transmitter placement for telecommunication to choose a few out of many examples has been observed To satisfy this demand a variety of real scenes cities suburban areas etc containing complex buildings have to be acquired Performing this task virtually saves time and replaces the need for physical models Our goal is the D reconstruction of complex buildings from a multitude of images Complex buildings have a polygo nal ground plan and a detailed roof shape Thus they are in general combinations and variations of the basic building types as shown in Figure Higher degrees of details contain also canopies dormers oriels chimneys and overhanging eaves One Eye Stereo System Englert G ulch cf Figure In this context we regard acquisition as a structured topological and geometrical modeling process using background knowledge A variety of ap Figure Some basic building types with di erent roof shapes at roof pent roof lean to roof saddleback roof gable roof hip roof broach roof pavilion roof hipped gable roof mansard roof gambrel roof proaches for the recognition of objects already exists SFH To reconstruct D objects current methods of Analytical and Digital Photogrammetry laser scanning and digital surface model analysis have to tackle with several problems occlusions inverse mapping and noise BKL The latter both methods are currently com plete inadequate to acquire a detailed D building description due to their limited resolution But they can support other approaches providing estimates to them For very moderate requirements on the complexity of D building structures BDR i e ground plan as polygon and one height for the building the operational but fully interactive Analytical Photogrammetry is a good solution With increasing demands on the degree of detail the e orts on the modeling are growing enormously for extended scenes To make this task tractable a trained operator is needed In contrast to these approaches Digital Photogrammetry o ers the possibility to au tomate the modeling process in partial and thus enables untrained operators to perform measurements Previous work in this area has amongst others been per formed by LS We adopted and substantially extended their approach using the Constructive Solid Geometry CSG for the modeling of complex buildings In Section the system HASE is described and discussed The modeling pro cess is done in monocular mode one eye stereo assisted by various supporting and automating tools for the form and pose adaptation of a large amount of CSG prim itives Our experiences on an extensive real scene are evaluated and compared with earlier experiments in Section Finally we conclude and give an outlook for future work Section