[visionlist] JOB: PostDoc - Multi-Source Data Registration for Dense 3D Urban Modeling

[TOURNAIRE Olivier]

Context
The IMAGINE (ENPC/CSTB) and MATIS (IGN) laboratories own a large collection of images acquired on dense urban areas. The wide variety of acquisition systems involved in the imaging process enables to acquire images ranging from low to very high resolution. A set of SLR cameras in a sphere (IMAGINE) mounted on a terrestrial vehicle as well as the Stereopolis system (MATIS), also equipped with laser sensors, provide georeferenced data with a high level of details in dense urban areas at street level (e.g. Google StreeView); images taken from helicopters or airships provide a higher coverage at medium resolution, also bringing information on human structures hidden from the ground (roofs…); large areas can be covered with aerial views with varying resolutions, from ten centimeters, to around a meter.
More than a decade of research in the field of 3D urban modeling and reconstruction in the both laboratories leads on the one hand, to the development of robust algorithms among the best in the world and, on the other hand, an ability to work with aerial images in a production framework for large scale area processing (BATI3D database at IGN).
Objectives
In the scope of the “Digital City” project, we would like to investigate the potential for joint use of data at different scales for the modeling of dense urban environments. The contribution of multi-source data and different view angles will also be taken into account in its ability to provide comprehensive and detailed reconstructions of all elements of buildings (roofs, facades, courtyards...).
Initially, we will therefore investigate the issue of matching multi-source data (different sensors and different points of view) and/or multi-temporal. The candidate will first compile a state of the art of the different methods and tools concerned with this problem by studying existing work in the computer vision and photogrammetric communities.
In a second step, the candidate will propose a processing framework integrating all available data, implement it on the basis of already existing tools, and evaluate it on a representative area. The chosen methodology will have to take into account the heterogeneous nature of data which particularly impact the matching step. Robust tools for filtering and matching interest points (e.g. SIFT) will thus be a major objective, as well as fitting and registering 3D point clouds.


References
D. Craciun<http://recherche.ign.fr/labos/matis/cv.php?prenom=&nom=Craciun>, N. Paparoditis<http://recherche.ign.fr/labos/matis/cv.php?prenom=&nom=Paparoditis> et F. Schmitt, 2010. Multi-view Scans Alignment for 3D Spherical Mosaicing in Large-Scale Unstructured Environments. Computer Vision and Image Understanding (CVIU), Accepted March 2010.
Labatut, P., Pons, J.-P. et Keriven, R., 2007. Efficient Multi-View Reconstruction of Large-Scale Scene Using Interest Points, Delaunay Triangulation and Graph Cuts. International Conference on Computer Vision.
Strecha, C., von Hansen, W., van Gool, L., Fua, P. et Thoennessen, U., 2008. On Benchmarking Camera Calibration and Multi-View Stereo for High Resolution Imagery. IEEE Conference on Computer Vision and Pattern Recognition.
VU, H. H., Keriven, R., Labatut, P. et Pons, J.-P., 2009. Towards High-Resolution Large-Scale Multi-View Stereo. IEEE Conference on Computer Vision and Pattern Recognition.
IMAGINE web site at École des Ponts ParisTech (ENPC): http://imagine.enpc.fr/
MATIS web site: http://recherche.ign.fr/labos/matis/
Required skills

·         The candidate should have a PhD degree in image processing, computer vision, computational geometry or 3D modeling

·         Good knowledge in mathematics for image, numerical optimization and 3D geometry

·         Proficiency in C++ is mandatory

Conditions

·         Duration: 12 months

·         Location: IMAGINE/ENPC (Champs-sur-Marne) and MATIS/IGN (Saint-Mandé), close to Paris

Contacts
Applications consist of a cover letter describing how your research experience is relevant to the proposed position, a resume and publications list. They should be sent to:


R. Marlet (IMAGINE / ENPC)

renaud.marlet at enpc.fr<mailto:renaud.marlet at enpc.fr>

+33 1 64 15 21 86


O. Tournaire (IMAGINE / CSTB)

olivier.tournaire at cstb.fr<mailto:olivier.tournaire at cstb.fr>

+33 1 64 68 83 23


M. Pierrot-Deseilligny (MATIS / IGN)

marc.pierrot-deseilligny at ign.fr<mailto:marc.pierrot-deseilligny at ign.fr>

+33 1 43 98 83 92

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://visionscience.com/pipermail/visionlist/attachments/20110225/974b1665/attachment-0001.htm>