TUTORIAL
Bassel Solaiman (Télécom Bretagne, Brest, France)

Basel Solaiman is a Full Professor at Telecom Bretagne / Telecom Institute in France. He is currently the Head of Information Processing Department at Telecom Bretagne in Brest France. His research’s activities include: Information and Data Fusion, Pattern Recognition, Image and Scene Interpretation and Signal processing. These activities have as major applications: Remote Sensing, Sonar Imaging, biomedical decision systems and Medical Imaging. He Chaired and Co-Chaired several International Conferences, is a guest editor and reviewer for a number of scientific journals, a visiting professor at several universities worldwide (Canada, USA, Poland, Tunisia, Hungry, Turkey, Syria). He has published several academic books and avec that 200 technical papers in international journals and conferences. In 2009, Pr. Solaiman has received the highest academic distinguishing French medal of “Cavalier of Academic Palms” for his academic and research contributions. 

Tutorial Abstract:

Information Fusion Theory and Pattern Recognition

The concept of Information Fusion was initiated by the need of using several data sources and technical terms like data fusion and multisensor fusion were born. This concept intends to give a global framework to pattern recognition systems using several sensors. Some important examples raised rapidly: Target Identification Systems, Parcels Recognition in remote sensing, Medical Diagnostic Systems, etc.

The use of existing theories like Bayesian, Belief Functions, Fuzzy and Possibility and the real world applications have certainly stimulated new research activities and new engineering challenges.

For instance: available information is not limited to simple data but includes knowledge sources (not necessarily numerical); different information sources may be affected by different forms of imperfections (missing data, uncertainty, ambiguity, etc.). These questions show clearly that information fusion is not the simple concatenation of exiting theories and that we need a global and homogeneous framework where:

• the basic concept of information is clearly defined;
• the mathematical imperfections models are defined and linked to imperfection sources affecting the observed, or available information;
• different existing theories and their scope of use and application are clearly positioned.        

This invited talk will tackle these important questions. A precise definition of the concept of an information element is given; different sources of information imperfections are presented, and, mathematical characterization of information elements is expressed so that the different information fusion technical approaches are positioned. Pattern recognition systems mainly using imaging sensors are used to clarify and give concert examples of the proposed global approach leading to consider Information Fusion as a single theory where existing technical approaches constitute the components of this theory.

Tutorial content:

1. Information Fusion Fundamental concepts
2. Information Elements and Imperfection sources in Pattern Recognition Systems
3. Fundamental Information Fusion Approaches:
    3.1. Bayesian Approach
    3.2. Belief Theory Approach
    3.3. Fuzzy Set Theory
    3.4. Possibility Theory
4. Application Examples