Skip to Main content Skip to Navigation

Simultaneous localization and mapping in 3D environments with stereovision

Abstract : In this thesis, we tackle the SLAM problem for robots evolving in 3D in large environments, without any prior knowledge on the environment, and using stereovision. A full implementation of the various processes has been conceived, developed, and experimented in various contexts. The first part of the thesis deals with the data association problem: it introduces a matching algorithm for invariant image point features, which is robust to image noise and viewpoint changes. The second part of the thesis is devoted to an implementation of a SLAM approach using an extended Kalman filter. Landmarks are the point features detected in the image, their 3D coordinate being computed by stereovision. The last part of the thesis presents and analyzes results with several hundreds meter long trajectories, with a low altitude flying blimp and a rover. When the trajectory ``closes a loop'', a fast backward correction of the robot poses based on the local motion estimates and the landmark map is applied, in order to be able to construct spatially consistent large digital elevation maps of the environment.
Document type :
Complete list of metadata

Cited literature [59 references]  Display  Hide  Download
Contributor : Emilie Marchand <>
Submitted on : Thursday, September 22, 2005 - 3:26:01 PM
Last modification on : Thursday, June 10, 2021 - 3:03:39 AM
Long-term archiving on: : Friday, April 2, 2010 - 10:20:47 PM


  • HAL Id : tel-00010250, version 1


Il Kyun Jung. Simultaneous localization and mapping in 3D environments with stereovision. Automatique / Robotique. Institut National Polytechnique de Toulouse - INPT, 2004. Français. ⟨tel-00010250⟩



Record views


Files downloads