Three-dimensional adaptive sensing of people in a multi-camera setup

Abstract

Sensing the presence and state of people is of paramount importance in assistive living environments. In this paper we utilise a set of fixed, calibrated cameras to model the bodies of people directly in three dimensions. An adaptive foreground segmentation algorithm is run per camera, providing evidence to be collected in 3D body blobs. A particle filter tracker allows monitoring the modelled bodies across time, offering estimations of their state by using hot-spots and body posture. We apply our system on fall detection and activity monitoring for the elderly, addressing both emergency and cognitive care.