Markerless visual tracking and motion analysis for sports monitoring
File(s)
Author(s)
Pansiot, Julien
Type
Thesis
Abstract
In the past decade, detailed biomechanical motion analysis has become an important part of athletic training and performance evaluation. However, most commercially available systems are obtrusive and require complicated experimental setup and dedicated laboratory settings. With recent advances in smart vision sensors, markerless vision-based approaches have attracted significant interests for detailed sport motion analysis as they do not involve the placement of external fiducials, thus providing pervasive measurement of the athletes without affecting their normal performance. This thesis presents a robust real-time vision-based tracking system based on a miniaturised, low-power, autonomous Visual Sensor Network (VSN). The system is able to provide real-time motion monitoring with on-node processing. Detailed technical issues concerning background and silhouette segmentation, canonical view generation, 3D model-based motion parametrisation and reconstruction, wearable-ambient sensor fusion and extraction of biomechanical motion indices are addressed. The proposed method is applied to motion tracking of indoor tennis training and performance evaluation. Ambiguities due to occlusion, view point dependency and a lack of depth information are reduced by the deployment of a VSN. A method is proposed to derive 2D canonical views from a set of input video sequences to facilitate consistent motion monitoring. Further steps for 3D model reconstruction and parametrisation are also proposed, which involve the use of spherical harmonic parametrisation for deriving a compact 3D shape descriptor. The remaining uncertainties in motion analysis are resolved by predictive tracking based on a biomechanical model such that issues related to occlusion are avoided. To further incorporate high-frequency biomechanical information such as ground reaction forces, a sensor fusion framework based on an integrated use of wearable and vision-based ambient sensors is proposed. The practical value of the proposed framework is demonstrated with a systematic implementation of a VSN for tennis training, which provides realtime automated generation of player motion profiles.
Version
Open Access
Date Awarded
2009
Copyright Statement
Attribution NoDerivatives 4.0 International Licence (CC BY-ND)
Advisor
Yang, Professor Guang- Zhong
Publisher Department
Department of Computing.
Publisher Institution
Imperial College London
Qualification Level
Doctoral
Qualification Name
Doctor of Philosophy (PhD)