Dr. Bentabet’s research is concerned with the spatiotemporal modeling of the environment (segmentation, detection, and tracking) using multiple cameras. Scene modeling with multiple cameras is a relatively new problem in computer vision, but one that has gained increasing interest recently. This comes from the fact that the presence of multiple cameras is necessary not only to help solve hard computer vision problems but also to provide the conceptual framework in which 3d reconstruction is possible. Dr. Bentabet is especially interested in modeling the sensors fusion and decision making mechanisms in situations where sensors produce incomplete and ambiguous description of the scene. These situations are usually caused by high occlusion, changing illumination, and camera calibration problems. Dr. Bentabet will focus on the following problems:
- Extraction of robust and reliable cues for objects modeling;
- Combination of the extracted cues from each sensor into a single and overall representation using notions of Dempster-Shafer theory;
- Development of robust decision making rules and quantification of the uncertainty.
The results of Dr. Bentabet’s project will provide the computer vision community with strong and alternative solutions to the sensors fusion problem. In addition, it will spur technology transfer to commercial applications, such as videoconference systems, surveillance, and activity monitoring.