Particle filter to track multiple people for visual surveillance

• Published in: IET computer vision Vol. 5; no. 4; pp. 192 - 200
• Main Authors: SHERRAH, J, RISTIC, B, REDDING, N. J
• Format: Conference Proceeding Journal Article
• Language: English
• Published: Stevenage Institution of Engineering and Technology 01.07.2011; John Wiley & Sons, Inc
• Subjects: Adaptation; Applied sciences; Benchmarking; Color; Computer vision; Detection, estimation, filtering, equalization, prediction; Exact sciences and technology; False alarms; Information, signal and communications theory; Signal and communications theory; Signal, noise; Surveillance; Telecommunications and information theory; Tracking; Visual; Target tracking; Birth death process; Background; Video technique; Labelling; Particle tracks; Modeling; Particle filter; Color image; Remote sensing; Remote supervision; False alarm rate
• ISSN: 1751-9632; 1751-9640
• DOI: 10.1049/iet-cvi.2010.0026

A particle filter (PF) has been recently proposed to detect and track color objects in video. This study presents an adaptation of the PF to track people in surveillance video. Detection is based on automated background modeling rather than a manually generated object color model. Furthermore, a labeling method is proposed to create tracks of objects through the scene, rather than unconnected detections. A methodical comparison between the new PF tracking method and two other multi-object trackers is presented on the PETS 2004 benchmark data set. The PF tracker gives significantly fewer false alarms owing to explicit modeling of the object birth and death processes, while maintaining a good detection rate.