Caroline Lacoste

How to contact me ?

INRIA - UR Sophia Antipolis
  • Phone : + 33 (0) 4 92 38 75 67
    		•
    2004, route des Lucioles, BP 93
  • Fax :+ 33 (0) 4 92 38 76 43
    		•
    06902 Sophia Antipolis- FRANCE 
  • E-mail :Caroline.Lacoste@sophia.inria.fr
    		•

    Research

    Resume
     

       Ma page en francais

    PhD subject:

             Extraction of Line Networks in Remote Sensing by Markov Object Process.
     

                With a view to cartographic update assistance, we aim to extract line  objects from satellite and aerial images.  The sought-after cartographic item consists of the different line networks such as hydrographic network, road network or subsurface network. This study is realized with the collaboration of French Geological Survey - BRGM, in french - (Nicolas Baghdadi, Chritine King). The treated sites will be chosen in connection with its main projects (SIG Africa: sites in equatorial Africa such as Guinea, on which optic imagery alone could not answer to cartographic preoccupations, and in Middle East where long wave radar data  may provide subsurface information, as for instance in Egypt).  The BRGM has large image data base on these sites as well as validation data.

                The approach is to build a Markov object process  in order to model the road network present in the image. Indeed, these recently developed models provide the same type of stochastic properties as those of Markov fields, while incorporating strong geometric constraints. These models allow to manipulate geometric objects, simple but generic due to characterization by variable parameters.  Interactions between objects are taken into account in the density of the process, which allows to incorporate constraints on the network topology, such as continuity or slight curvature. Data properties are taken into account in the density of the process through a data term based on statistical tests. Optimization is realized by simulated annealing using a RJMCMC algorithm.
     

    Publications:

    Communications: