The talk will give an overview of the "structure from motion" problem, i.e. how in Computer Vision the camera network parameters are automatically computed simultaneously to the 3D object coordinates of tie points. Some methodologies are common (e.g. bundle adjustment), while other distinctive and innovative solutions to the problem were developed in the Computer Vision community and adopted by the Photogrammetric community in order to speed up the entire image-based pipeline. Examples and open issues will be also discussed.
Prof. Andrea Fusiello
Andrea Fusiello received his Laurea (Master) degree in Computer Science from the University of Udine (Italy) in 1994. In 1993-94 he worked within the Computer Vision Group of FBK Trento and he received his PhD in Computer Engineering from the University of Trieste in 1999. He had been Visiting Research Fellow at Heriot-Watt University, Edinburgh in 1999. From 2000 to 2011 he has been with the the Dept. of Computer Science, University of Verona (Italy), where he taught Computer Graphics and Computer Vision. In 2012 he moved to the University of Udine, at the DIEGM Dept., where he teaches Fundamentals of Computer Science (undergraduate) and Computer Vision (graduate). He has been Associate Editor of "Pattern Recognition" Journal for three years and now he is A.E. of IET Computer Vision. He regularly serves on the Program Committee of CVPR, ICCV and ECCV Conferences. His research focuses on various topics in Computer Vision and Image Analysis: stereo matching, epipolar rectification, tracking, motion segmention, mosaics, view synthesis, auto-calibration, 3-D modelling, model acquisition. He is chairman and co-founder of 3Dflow srl, a spinoff of the University of Udine.
Prof. Andreas Georgopoulos
Documenting the Cultural Heritage, be it tangible or intangible, is a necessity nowadays. During the recent years, with the development of ICT techniques, this task has experienced a tremendous progress. However Cultural Heritage documentation involves the cooperation of Users (i.e. archaeologists, architects, curators) and Providers (i.e. ICT specialists, surveyors, engineers, photogrammetrists, etc.). At the same time the adaptation and practical implementation of these advances is met with some skepticism and confronts difficulties because of (a) tradition, (b) lack of information and “fear” of the unknown from the part of the archaeologists and (c) lack of the necessary sensitivity from the part of the ICT and Geomatics experts. Hence it is of utmost importance to advance the familiarization of all future experts to these contemporary techniques, to the identification and understanding of Users' needs combined to available tools for the Providers for serving restoration, preservation and management actions. This can only be achieved through interdisciplinary actions at a theoretical and practical level. The talk will brief reference the latest developments in data acquisition methods and technologies, data processing methodologies and contemporary documentation products. Furthermore examples of good (and bad?) practices of Users/Providers cooperation are given with a future outlook. The role of CIPA (The International Committee for Documentation of Cultural Heritage) in this matters is also identified.
Andreas Georgopoulos (Director of the Laboratory of Photogrammetry) received the Diploma in Rural and Surveying Engineering from the National Technical University of Athens (NTUA) in 1976, the Master of Science (MSc) and the PhD in Photogrammetry from University College London in 1977 and 1981, respectively. Since 1985 he is a Faculty member of the School of Rural & Surveying Engineering of NTUA. Since 1999 he has been elected Full Professor in NTUA and from 2002 to 2006 he has served as Head of the School of Rural & Surveying Engineering of NTUA. From 2010 to 2011 he has served as Visiting professor at the Cyprus Institute and the Cyprus University of Technology and since 2009 is also visiting Professor at the Postgraduate program of Monument Conservation of RLICC (Catholic University of Leuven). Since 1985 he has participated in all research projects of the Laboratory of Photogrammetry of NTUA involved with Photogrammetry, Architectural Photogrammetry, Digital and Analytical Photogrammetry, Cadastre and Geomatics in general. Since 1996 he has been National Delegate for Greece in CIPA-Heritage Documentation. He was Director of 2007 CIPA International Symposium and he is a member of the CIPA Executive Board since 2005, serving since 2011 as General Secretary.
Dr. Sven Havemann
BIM (Building Information Modeling) is currently gaining massive momentum, and IFC (Industry Foundation Classes) appears to become the main building description standard for many years to come. But BIM is more than points or triangles: rich semantics is the key to its success. This talk will highlight the challenges in moving from sampled to semantically rich shape descriptions, and explain why (and how) procedural shape modeling can be of great help in this transition - and beyond. This is illustrated by practical examples using GML (Generative Modeling Language).
Sven Havemann has some 15 years of experience in research on generative shape modeling. He graduated from Bonn university in 1997 ("Generative Modeling") and got a PhD with distinction from TU Braunschweig ("Generative Mesh Modeling") where he developed GML. Dr Havemann is leading a research group on geometric modeling at TU Graz, Austria.
Dr. Florent Lafarge
In the first part of this talk, I will present a literature review on 3D reconstruction from multi-view stereo images and its applications 3D architectural and urban modeling from ground or aerial acquisitions. We will see how the quality of a 3D model can be measured, and also how Computational Geometry can help traditional photogrammetric algorithms to generate geometrically accurate, compact and structure-aware meshes. In the second part of this talk, I will focus on recent methods for reconstructing objects and scenes by hybrid surfaces combining free-form parts and canonical parts generated from geometric primitives.
Florent Lafarge is a research scientist at INRIA Sophia Antipolis, France. He works within the Titane research group on the geometric modeling from physical measurements. He received the MSc degree in applied mathematics from the University of Toulouse, France, and the PhD degree in applied mathematics from the Ecole des Mines ParisTech, in 2007. He has been a postdoctoral researcher at the University of Auckland, New Zealand and at the Imagine group of Ecole des Ponts ParisTech. His research interests focus on urban scene analysis, surface reconstruction and probabilistic modeling in vision.