Selected Readings in Vision and Graphics
edited by Luc Van Gool, Gábor Székely, Markus Gross, Bernt Schiele

Volume 40

Martin Spengler
On the Applicability of Sequential Monte Carlo
Methods to Multiple Target Tracking

First edition 2005, 168 pages, € 64,00. ISBN 3-86628-021-1

This book provides a systematic analysis of different attempts to multiple target tracking  using Sequential Monte Carlo (SMC) methods. In recent years, SMC methods - also known as  Particle Filters - have gained increasing popularity in the context of visual object  tracking. They provide simple yet powerful techniques for tracking objects efficiently in  the presence of clutter and occlusion. Though SMC approaches have been applied successfully to the single object case, their ability to track multiple targets is still subject to ongoing research.
In this work, the applicability of SMC methods to multiple target tracking is evaluatedby  means of the quantitative as well as qualitative analysis of the performance achieved by SMC based algorithms. Three principal SMC based multi-object tracking strategies are identified, each representing a particular class of particle-based algorithms for  probabilistic multiple target tracking. In order to minimize the effects arising from object models and video footage, the tests are conducted using various state-of-the-art object models and a variety of different video sequences.
In addition, the three principal tracking strategies are compared with BraMBLe, a highly  renowned SMC based people tracker. In contrast to the algorithms tested in this work, BraMBLe uses a three-dimensional model for tracking people. The quantitative and qualitative comparison attempt to answer the question whether or not true  three-dimensional models are worth the additional costs arising from their increased complexity.

Martin Spengler obtained a M.Sc. degree in Computer Science (Dipl. Ing.) from the Swiss Federal Institute of Technology (ETH),Zurich, Switzerland in 2000. From 2000 to 2005, he has been research assistant and Ph.D. student at the Perceptual Computing and Computer Vision Group of the Swiss Federal Institute of Technology (ETH), Zurich, Switzerland. He earned his Ph.D. in Computer Science (Dr. sc. techn.) for his work on the applicability of Sequential Monte Carlo Methods to the vision-based multiple target tracking problem.

Keywords: Multiple Target Tracking, Sequential Monte Carlo Methods, Performance Evaluation, Computer Vision

Reihe " Selected Readings in Vision and Graphics " im Hartung-Gorre Verlag

Direkt bestellen bei / to order directly from: Hartung.Gorre@t-online.de