Silhouette Based Real Time Human Motion Tracking

INVESTIGATORS

Dr. Yisong Chen et al.

 

KEYWORDS:

Multiple view reconstruction, Epipolar geometry, Camera calibration, Motion analysis, Feature extraction, Real-time tracking, Pose estimation.

BRIEF DESCRIPTION

Compared with previous approaches, our human motion tracking framework has following advantages

1, We put least requirement on environment. As few as only two cameras are used to achieve satisfactory human motion tracking. An approximate upright posture and a flat ground are the only two assumptions.
2, This is the first attempt of making full use of one top view camera. The top view camera provides special superiority in many aspects, such as calibration calibration, motion tracking and pose recovery.
3, This is a true-sense real time framework. The real time performance is beneficial from the fact that most processing is executed in one-dimensional context. An average speed of 25 frames per second is reached in an ordinary PC workstation. This is a great improvement compared to results of previous work.
4, The human motion is not as strictly constrained as many related work. Orientation-free walking motion can be effectively tracked with satisfactory ambiguity solution.
 

Well, following are several of our silhouette based human motion tracking results. The tracking can be performed in real time with an average frame rate of 25 frames/second.

           Figure 1. Walking motion tracking

            Left: side view, right: top view

             Figure 2. Arm motion tracking

            Left: side view, right:top view

 

FUNDING AGENCY

N/A

 

PUBLICATIONS

N/A