Visual Person Tracking Using a Cognitive Observation Model

In this article we present a cognitive approach to person tracking from a mobile platform. The core of the technique is a biologically inspired observation model that combines several feature channels in an object and background dependent way, in order to optimally separate the object from the background. This observation model can be learned quickly from a single training image and is easily adaptable to different objects. We show how this model can be integrated into a visual object tracker based on the well known Condensation algorithm. Several experiments carried out with a mobile robot in an office environment illustrate the advantage of the approach compared to the Camshift algorithm which relies on fixed features for tracking.