Index
 
  • Visual Servoing
    • Uncalibrated 2 1/2 D visual servoing
    • Visual servoing with a zooming camera
    • Visual servoing of non-holonomic robots
  • Active Vision
    • Robust 3D Reconstruction from stereo
    • Estimation of the depth distribution without camera calibration
  • Computer Vision
    • Euclidean reconstruction without direct camera calibration
    • Structure from motion without model selection
    • Images matching of unknown planar contours
    • Camera self-calibration from unknown planar structures
  • Robot Control
    • Robot control from disparate multiple sensors
    • Control of a redundant manipulator arm in degenerate directions
 
 
Collaborators
 
Selim Benhimane: PhD 2003-2006 (research associate at the Technische Universitat Munchen, Germany)
 
Vincent Brandou: PhD student
 
Geraldo Silveira: PhD student
 
Andrew Comport : Research associate 2005-2007 (research scientist at CNRS Clermont-Ferrand, France)
 
Visual Servoing

Uncalibrated 2 1/2 D visual servoing
 

My research work concerns robust vision-based robot control. During my Ph.D., I worked in the Vista project of the IRISA/INRIA Rennes on a new visual servoing technique which allows to avoid the disadvantages of the classical position-based (3D) and image-based (2D) visual servoings. This new method, called 2 1/2 D visual servoing, presents the followings advantages:

•  a convergence domain as large as the task space;
•  a large robustness with respect to calibration errors;
•  a minimum a priori information on the structure of the observed objects.

The main application of this work is the control of the six end-effector d.o.f. of a robot for any tracking or positioning task. More particularly, this work was supported by the national French Company of Electricity Power (EDF) in order to realize the automatic manipulation of maintenance tools in its nuclear power plants.


Computer Vision

Matching of planar contours for visual servoing
 

The aim of this research is to design a complete system for segmenting, matching and tracking planar contours for use in visual servoing. Our system can be used with arbitrary contours of any shape and without any prior knowledge of their models. The system is first shown the target view. A selected contour is automatically extracted and its image shape is stored (Mpeg 0.5 Mb). The robot and/or object are then moved (Mpeg 0.4 Mb) and the system automatically identifies the target (Mpeg 0.5 Mb). The matching step is done together with the estimation of the homography matrix between the two views of the contour. Then, a 2½D visual servoing technique is used to reposition the end-effector of a robot at the target position relative to the planar contour (Mpeg 1.0 Mb). See the complete example with two contours of general shape (Mpeg 2.5 Mb). The system has been successfully tested on several contours with very complex shapes such as leaves, keys and the coastal outlines of islands.

The main application of this system in the VIGOR project is the visual servoing with respect to contours on the part of a ship. First, the reference contour is selected (Mpeg 0.4 Mb). Then the robot is moved to its initial position (Mpeg 0.5 Mb). The image is segmented in order to obtain the contours. The reference contour is automatically matched with the current one Mpeg 0.7 Mb). Finally, the 2½D visual servoing technique is used to reposition the end-effector of the robot (Mpeg 1.4 Mb). See the complete sequence of the experimental results (Mpeg 3.2 Mb).


Active Vision


The aim of this future research work is to improve the reconstruction of 3D object using controlled camera motions.


Robot Control


Robot control from disparate multiple sensors
 

...


Control of a redundant manipulator arm in degenerate directions