Accès direct au contenu


Version anglaise


Accueil > Publications > Thèses du CMLA

Earth Observation and Stereo Vision

le 5 octobre 2015

Thèse de Carlo De Franchis (CMLA)



This thesis deals with the problem of computing accurate digital elevation models of the Earth's surface from optical images taken by pushbroom observation satellites. It takes advantage of the collaboration of the defendant with CNES (the French Space Agency) on the development of stereo vision tools for Pléiades, the first Earth observation satellite producing quasi simultaneous stereo pairs or triplets with small baseline.

The first chapter describes a simple pushbroom camera model for observation satellites orbiting around the Earth and addresses the correction of the acquisition geometry by involving extrinsic information. This chapter proposes a new algorithm to refine the orientation parameters from a set of ground control points, applicable to all pushbroom satellites.

With the goal of testing for satellite imaging the thriving exploration of stereo matching by the computer vision community, the second chapter explores the adaptation of the theory of epipolar resampling to pushbroom images. Epipolar resampling is traditionally used in stereo to reduce the matching computational cost, and permits to test for satellite imaging the most competitive computer vision algorithms. The third chapter discusses the effects of geometric calibration inaccuracies and proposes a method to cancel its impact on stereo matching.

The fourth chapter analyzes and describes a detailed implementation of the Semi-Global Matching (SGM) algorithm, which is currently among the top-ranked stereo vision algorithms. Based on a recently proposed interpretation of SGM as a min-sum Belief Propagation algorithm, a variant is proposed that allows to reduce by a factor five the energy gap of SGM with respect to reference algorithms for Markov Random Fields with truncated smoothness terms.

By wrapping together the algorithmic blocks described in the previous chapters, the fifth chapter describes S2P, a complete stereo pipeline for producing digital elevation models from satellite images. As an application, a landscape evolution model is presented in the sixth chapter. The model is used to simulate numerically the fine structure of the river networks on digital elevation models obtained from Pléiades Earth observation images.

The source code of the S2P stereo pipeline is distributed as open source. To ensure reproducibility, the algorithms implemented in each step of the S2P pipeline are submitted to the IPOL journal, with detailed descriptions of the algorithms, documented source codes and online demonstrations for each block of the pipeline.

Type :
Thèses - HDR
Lieu(x) :
Campus de Cachan
Bâtiment Laplace, Salle Renaudeau, R-de-Ch.

Jury de thèse

  • Jean-Michel Morel
  • Gabriele Facciolo
  • Pascal Monasse,
    Ecole des Ponts ParisTech (referee)
  • Marc Pierrot Deseilligny,
    IGN (referee)
  • Hélène de Boissezon,
    CNES (examiner)
  • Pablo d'Angelo,
    DLR (examiner)
  • Jean-Marc Delvit,
    CNES (examiner)
  • François Goulette,
    Mines ParisTech (examiner)
  • Julien Michel (invited)

Prix du meilleur article étudiant 2014 pour Carlo De Franchis

L'article "Automatic Digital Surface Model Generation From Pleiades stereo Images" de Carlo de Franchis [et al.], publié dans le n°208 de la Revue Française de Photogrammétrie et Télédétection (SFPT) est consacré aux "Pléiades Days" du CNES. EN savoir plus...

Recherche d'une actualité

Recherche d'une actualité