[Imageworld] PhD Student Position In Computer Vision at Université de Bourgogne, France

[desire sidibe]

PhD Proposal:  Vision based robot navigation: a visual
attention based approach.




The 3D vision team of the LE2I laboratory at
Université de Bourgogne, Le Creusot in France, and the Hubert Curien Laboratory
at Université de St-Etienne in France, are seeking a highly motivated PhD candidate
in the field of Computer Vision and Image Processing. The PhD student will work
in a multi-disciplinary and multi-cultural environment between both
laboratories.

 

Context------------


Many computer vision applications require the
detection, localization and tracking of regions/objects of interest in an image
or a video.

Biologically motivated approaches that
automatically find regions of interest based on concepts of the human visual system
have proved to be effective solutions in different applications. However, most
of existing techniques are limited to the analysis of static images and few
methods exploit the temporal components of video sequences.

 

In the context of autonomous navigation, a
mobile robot needs to locate and move itself in uncontrolled environments. A
camera can be used as principal or unique sensor to achieve autonomous
navigation through the task of SLAM (Simultaneous Localization and Map building applications). However, a key requirement in
visual SLAM is the ability to detect useful landmarks which are easy to track
and stable over time [1].

 

Objectives-------------


The goals of the thesis are twofold:

1.      Firstly, we aim at analysing
different proposed saliency models and develop a spatio-temporal saliency model
based on human visual speed perception [2, 3].

2.      Secondly, the saliency model will be
used for a mobile robot navigation problem. This includes answering the
following question: 

-         
How
does an attention/saliency model can help in robot’s self localization?

-         
How
can an attention/saliency model be used for automatic landmark selection and
tracking is visual SLAM?

-         
How
does a motion segmentation model can be used to detect/track salient moving
object from background?

 

Special attention will be given to non-deterministic
approaches (e.g. particle filters models) and to the computational complexity
of the developed approach. 

The algorithms will be tested on a mobile robots
platform.

 

Application--------------


Candidates should have a strong mathematical
and statistical background, with an MSc or equivalent degree in one of the
following fields: Computer Science, Electrical Engineering, Mathematics, and
Physics.



Knowledge of
Signal/ Image Processing is highly desirable. Knowledge of Robotics is useful.



Candidates
should have excellent programming skills (eg. Matlab, C/C++) and should be
fluent in English. 


The successful
candidate is expected to start on September or October 2010, and complete his
PhD in 3 years.

 

Applicants
should send their CVs, letter of intentions and one or two recommendation letters
(with contact details) before June 30th to:  

Dr Désiré
Sidibé: dro-desire.sidibe [at] u-bourgogne.fr.Prof Alain Trémeau : alain.tremeau [at] univ-st-etienne.fr







Some References

----------------------- 


1. J. Davison, I. Reid, N. Molton, O. Stasse,
MonoSLAM: Real-Time Single Camera SLAM, in IEEE PAMI, 2007.

2. C. Liu,
P. C.Yuen, G. Qiu, Object motion detection
using information theoretic spatio-temporal saliency, Pattern Recognition,
2009.

3. Z. Wang and Q. Li, Video quality assessment
using a statistical model of human visual speed perception, JOSA, 2007.

 




      
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.diku.dk/pipermail/imageworld/attachments/20100604/4d76390f/attachment.htm