Texture Mapping of Non-Orthorectified Images onto Topographic Model

Claude Cavelius and Luc Buatois and Sophie Viseur and Guillaume Caumon. ( 2008 )
in: Proc. 28th Gocad Meeting, Nancy

Abstract

Whereas aerial and satellite pictures are often used when building 3D geological models from eld observations, simple landscape pictures are often neglected as a direct input to modeling. One reason is that geomodeling packages do not have the capability to georeference such pictures, which are distorted. Model quality would be really improved if it was given the possibility to take them into account: they provide textured views of the topographic surface, making it easier to identify geological features such as faults or layer boundaries. Also, such pictures can be taken routinely, and do not involve time-consuming acquisition procedure as with LIDAR images. Therefore, we propose a method to map eld pictures taken with a simple camera onto an existing Digital Elevation Model. This method retrieves the camera projection only from a few calibration points and the camera viewpoint. Special care is taken to address visibility issues, so that only the visible part of the topography is textured, and that hidden surfaces are left untouched. Once calibrated, our system also makes it possible to project 2D picture digitalizations onto the 3D model, giving thus the alternative to use them as further input to the 3D model. This technique may also be applied in the context of LIDAR acquisition, especially in the case of high cli s for which a helicopter survey is required, making it dicult to georeference the associated photos. This method has been tested on two case studies located in SE France: the turbiditic Annot Basin and the carbonate platform margin observed on the cli of Gresse-en-Vercors.

Download / Links

    BibTeX Reference

    @inproceedings{112_cavelius,
     abstract = { Whereas aerial and satellite pictures are often used when building 3D geological models from
    
    eld observations, simple landscape pictures are often neglected as a direct input to modeling. One
    reason is that geomodeling packages do not have the capability to georeference such pictures, which
    are distorted. Model quality would be really improved if it was given the possibility to take them
    into account: they provide textured views of the topographic surface, making it easier to identify
    geological features such as faults or layer boundaries. Also, such pictures can be taken routinely,
    and do not involve time-consuming acquisition procedure as with LIDAR images.
    Therefore, we propose a method to map 
    eld pictures taken with a simple camera onto an
    existing Digital Elevation Model. This method retrieves the camera projection only from a few
    calibration points and the camera viewpoint. Special care is taken to address visibility issues, so
    that only the visible part of the topography is textured, and that hidden surfaces are left untouched.
    Once calibrated, our system also makes it possible to project 2D picture digitalizations onto the 3D
    model, giving thus the alternative to use them as further input to the 3D model. This technique
    may also be applied in the context of LIDAR acquisition, especially in the case of high cli
    s for
    which a helicopter survey is required, making it dicult to georeference the associated photos.
    This method has been tested on two case studies located in SE France: the turbiditic Annot
    Basin and the carbonate platform margin observed on the cli
    of Gresse-en-Vercors. },
     author = { Cavelius, Claude AND Buatois, Luc AND Viseur, Sophie AND Caumon, Guillaume },
     booktitle = { Proc. 28th Gocad Meeting, Nancy },
     title = { Texture Mapping of Non-Orthorectified Images onto Topographic Model },
     year = { 2008 }
    }