Model-based people tracking-- probabilistic approach
|
We present a robust and effective model-based approach to recovering motion parameters of walking people from monocular image sequences in a CONDENSATION framework. From the semi-automatically acquired training data, we learn a motion model represented as Gaussian distributions, and explore motion constraints by considering the dependency of motion parameters and represent them as conditional distributions. Then both of them are integrated into a dynamic model to concentrate factored sampling in the areas of the state-space with most posterior information. To measure the observation density with accuracy and robustness, a PEF (Pose Evaluation Function) combining both boundary and region information is proposed. The function is modeled with a radial term to improve the efficiency of the factored sampling. We also address the issue of automatic acquisition of initial model pose and recovery from severe failures. A large number of experiments carried in both indoor and outdoor scenes demonstrate that the proposed approach works well. |
Result of indoor scene: video and selected images
Result of outdoor scene1: video and selected images
Result of outdoor
scene2: video and
selected images
Publications
1. H. Ning, T. Tan, L. Wang, and W. Hu, People Tracking Based
On Motion Model and Motion Constraints with Automatic Initialization, Pattern Recognition, 37(7), 1423-1440, 2004. [PDF].
2. Huazhong
Ning,
3. Huazhong Ning, Model-based Tracking of Walking People, Institute of Automation, Chinese Academy of Sciences, Master’s thesis, advisor: Prof. Tieniu Tan, 2003 (in Chinese) [PDF]