Calibration-free target detection based on thermal and distance sensor fusion

Abstract

Infrared (IR) thermal vision systems provide a passive and contact-less framework to evaluate temporal signatures of people presence in indoor scenarios. However, static 2D IR thermal projection of complex 3D objects cannot provide sufficient information for large-scale and continuous people estimation tasks. This paper proposes a change-point detection algorithm that jointly fuses thermal and distance information obtained from an IR array and an ultrasonic distance sensor to detect targets, namely human subjects, inside an indoor environment. An extensive validation phase has been carried out through experimental trials that have been conducted in a smart office using ceiling-mounted devices. Unlike previous works in this area, the proposed approach eliminates time consuming calibration steps by highlighting the benefits of the IR thermal and ultrasonic sensor fusion framework.