Place-Dependent People Tracking

People typically move and act under the constraints of an environment, making human behavior strongly place-dependent. Motion patterns, the places and the rates at which people appear, disappear, walk or stand are not random but engendered by the environment. In this paper, we learn a non-homogeneous spatial Poisson process to spatially ground human activity events for the purpose of people tracking. We show how this representation can be used to compute refined probability distributions over hypotheses in a multi-hypothesis tracker and to make better, place-dependent predictions of human motion. In experiments with data from a laser range finder, we demonstrate how both extensions lead to more accurate tracking behavior in terms of data association errors and number of track losses. The system runs in real-time on a typical desktop computer.