Model-based people tracking--kinematical approach
|
We present a simple but effective method for recovering 3D human body walking parallel to the image plane from monocular video sequences based on 3D human body model, using robust image matching and specific search strategy. Monocular 3D body tracking is challenging, partly because at least 40 joint parameters need to be estimated. These parameters are reduced to 12 in our work under the assumption of walking parallel to the image plane. To effectively measure the match error, a pose evaluation function combining both boundary and region match error is provided. Simultaneously, we calculate the physical forces that are used as heuristic information in subsequent search. Experiments with both indoor and outdoor scenes demonstrate that our approach has an encouraging performance. |
Results: video and selected images
Publications:
1. H. Ning, T. Tan, L. Wang and W. Hu,
Kinematics-based tracking of human walking in monocular video sequences, Image
and Vision Computing, 22(5), 429-441, 2004. [PDF]
2. Huazhong
Ning,
 |