Fouad Bousetouane
RTIS-Laboratory
University of Nevada Las Vegas
USA
Title: Uncertainty Quantification in Computer Vision Problem: Application to Transportation
Biography
Biography: Fouad Bousetouane
Abstract
Nowadays, Monitoring roadways to ensure safety of vehicles and pedestrians is a challenging problem because of the high volume of traffic. The development of intelligent and omnipresent systems for automatic monitoring of modern roadway becomes indispensable. With the technological advances in sensors design, communication, computer vision and distributed inferences are stimulating the development of new innovative and intelligent techniques that will help transportation agencies and enforcement officers to ensure safety and improve traffic flow. Visual sensor network technology is seen to play an important role in such application. However, the aggregation and the interpretation of distributed visual information in real-time is still a biggest challenge. The complexity of such operations is mainly caused by the presence of multilevel uncertainty. Uncertainty in trajectory estimation of vehicles, visual signatures of vehicle and pedestrian, travelling time across visual-sensors, poses, etc. This explosion of uncertainty will certainly affect the global decision of the automated roadway monitoring systems. The major question should be asked today is how we can quantify this explosion of uncertainty to improve the decisional process of the automated visual monitoring systems? Through this talk, I will attempt to answer this question through the presentation of new approaches that integrate a combination of multi-level distributed artificial intelligence, dynamic computer vision techniques and filtering theory.