I was a PhD student at theUniversity of Rouen between october 2005 and december 2008. I joined the Intelligent Transportation System team of the LITIS laboratory. Pr. Pierre Miché as well as Christèle Lecomte supervised my work . I am currently a non-permanent teacher and reasearch assistant at the University of Rouen.

My research activities are mainly dealing with:

- data fusion and the belief function theory (also called Dempster-Shafer theory)

- computer vision and visual tracking

- intelligent vehicles.

The goal of my PhD works is to enable a robust tracking of the preceding vehicle in on-road sequences of images. To reach this goal, we came to the conclusion that a one-source approach is insufficient to deal with all the disrupting events that occur in a natural scene like occlusions, clutter, illumination changes, fast movements or particular movements. Indeed a source of information corresponds in our case to a visual information extraction technique, and there is no technique that is insensitive to all the disrupting event hence the need for data fusion. By cleverly chosing several extractors, it is possible to overcome these difficulties and cumulate the properties of the extractors.

Particle filters are widely used algorithms for tracking applications. They can be run quicly and better resist to occlusions, therefore they were selected as the main tracking procedure to work on. Belief functions are also well-known for their ability to process data from several sources and offer an efficient information combination. Since belief functions can be transformed into probabilities using the Transferable Belief Model of Smets, the fused information can be incorporated into a particle filter.

Finally, since some contextual information can be obtained via the particle filter, a new combination pattern of belief masses can also be designed. The combination of these contributions produced promising results on on-road sequences of several kinds (different weather conditions, different road types, different traffic situations).


My research activities have resulted so far in three publications in international conferences:

[1] Klein, J.; Lecomte, C. & Miche, P.
Fast color-texture discrimination: application to car-tracking
IEEE int. conf. on Intelligent Transportation Systems ( ITSC), 2007, 541-546

[2] Klein, J.; Lecomte, C. & Miche, P.
Tracking objects in videos with texture features
IEEE int. conf. on Circuits Electronisc Cybernetics and Systems, 2007

[3] Klein, J.; Lecomte, C. & Miche, P.
Preceding car tracking using belief functions and a particle filter
IEEE int. conf. on Pattern Recognition, 2008, to appear

My Phd manuscript is also available:
J. klein - Suivi robuste d'objets dans les séquences d'images par fusion de sources, application au suivi de véhicules dans les scènes routières - University of Rouen, 2008.


Concerning teaching activities, I have also been teaching several subjects in the past few years: digital coding, digital electronics, probability theory and object oriented programming (c++ and Java).

I am also a member of the GdR ISIS and attended meetings on visual tracking or data fusion. I used to be an elected member of the laboratory council as well as an elected member of the department of physics council of the University of Rouen.

I have been part of the Chinese students and researcher union of Rouen.

I attended the summer school of Peyresq in 2006 which dealt with Monte Carlo methods.
I attended the winter school on digital color image processing in Poitier in 2007 (EHINC 07) and presented a poster.