3D Head Tracking using the Particle Filter with Cascaded Classifiers

                       Akihiro Sugimoto 

            National Institute of Informatics, Japan

Abstract:


In the last decade, a variety of tracking techniques based on
particle filters have been proposed.  However, most of the existing
methods adopt only simple perceptual cues such as color histograms
or contour similarities for hypothesis evaluation.  To improve the
robustness and accuracy of tracking, it is very important to have a
better means for hypothesis evaluation.  In this talk, I present a
novel hypothesis evaluation technique for human head tracking using
cascaded classifiers based on AdaBoost and Haar-like features. In
particular, we use multiple classifiers, each of which is trained for
detecting human heads of a particular direction.  Among these classifiers,
the most suitable one is selected adaptively by considering each hypothesis
and known camera position.  Our experimental results demonstrate the
effectiveness and robustness of our method.