Adaptive Object Tracking in Dynamic Environments with User Interaction
Gabriele Peters, Martin S. Kluger · 2009
Abstract—In this article an object tracking system is introduced which is capable to handle difficult situations in a dynamically changing environment. We evaluate the concepts of the proposed system by means of the task of person tracking in crowded seminar rooms. The contributions are threefold. We propose an expansion of the condensation algorithm which results in a more stable tracking in difficult situations such as sudden camera movements. Secondly, a new way of particle diffusion is introduced which allows for the adaption of the tracking module to movements of the camera as, e.g., induced by the user. These two contributions apply not only to person tracking but to object tracking in general. The third contribution, which applies to person tracking only, consists in a flexible way how to represent a moving person. The proposed tracking module meets demands such as real-time tracking of a target person who is allowed to move freely within a group of other persons and thus can be occluded. An interactive selection of new target persons is possible and also the challenge of a dynamically changing background could be coped with. Index Terms—person tracking, video surveillance, automatic video production, intelligent rooms I.