The Current State and TRL Assessment of People Tracking Technology for Video Surveillance Applications
Abstract
People tracking is a fundamental problem in computer vision and video surveillance. Automated person tracking is the foundations of several video analytics tasks that have a potential of significantly facilitate the efficiency of using video surveillance technology. This includes human activity/event recognition (for example, recognition of suspicious activities such as tailgating and loitering), scene analysis (for example, computing the number of visitors to a specific area), and calculation of face logs(for example, storing multiple views of the same person's face, for future analysis tasks such as face recognition). This paper presents a survey of techniques developed for tracking people in video, followed by the technology readiness level (TRL) assessment for people tracking technology based thereon. Two main approaches for the problem are identified and comparatively reviewed: one based on appearance models and the other based on motion models. The technology readiness assessment is performed for different surveillance-based scenarios of increasing complexity: at passport control kiosk, controlled chokepoint, uncontrolled chokepoint, little and dense traffic, and outdoors. For each surveillance setup type, the techniques that show the best reported performance are identified. Additional background related to people re-identification video and the object tracking techniques developed at the University of Ottawa is provided.
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 01, 2014
- Accession Number
- AD1018058
Entities
People
- Chris Whitten
- Diego Macrini
- Dmitry O. Gorodnichy
- Ghazal Moradian
- Robert Laganiere
- Vafa Khoshaein
Organizations
- Canada Border Services Agency