Human Motion Analysis
Research led by Prof. Adrian Hilton has introduced techniques for 3D reconstruction and tracking of human motion from single or multiple view video. Video-based reconstruction allows non-invasive acquisition of 3D human motion over large areas such as stadium sports (football, rugby) or athletics.
Reconstruction of 3D human motion from video allows the measurement of movement at a distance without a requirement for placement of markers on the subject. The aim is to measure full body motion in natural environments such as a sports stadium during game play for analysis of player performance and team interaction. This technology also has potential for clinical application in the measurement of walking gait and general movement for rehabilitation.
V-Lab research has introduced methods for reconstruction of 3D human skeletal motion and shape from multiple view video acquisition. 3D video acquisition enables the surface shape and motion of dynamic scenes to be captured. Research has introduced methods for the recognition and reconstruction of human pose from multiple view video. In a controlled indoor setting this allows recovery of detailed body pose and surface shape dynamics including cloth and hair movement. Recent advances in temporal surface alignment allow the tracking and analysis of surface motion together with the underlying skeletal motion.
In stadium sports played over a large-area our research has introduced techniques for recognition of 3D player pose. This allows the tracking of players and their movements in team sports such as football or rugby providing the basis for subsequent measurement and analysis.
V-Lab is collaborating with professionals in human movement and sports performance analysis, together with motion capture companies, to develop technologies for natural motion analysis from video. Research is also investigating methods for measurement and classification of body shape from video which has potential applications in clinical analysis of obesity and fashion retail.
- SCENE Novel Scene Representations for Richer Networked Media (EU IST)
- BodyShape Body Shape Recognition for Online Fashion (EPSRC)
- Analysis of Human Motion in Sports (BBC)