Modeling the Systematic Uncertainty of Photogrammetric Tracking of Human Motion
We address bias errors of photogrammetric tracking of four SELSPOT-II® cameras using active marker photogrammetry in a 2 m × 2 m × 2 m viewing volume for human locomotion measurements. We present uncertainty modeling regarding the first stage of equipment set up, which provides the camera frame to global frame rotation matrices and the distances among cameras. We also characterize the uncertainty due to the camera distortions of the bare system as compared to published performances achieved with a camera correction procedure. The particular approach is to qualify performances of photogrammetric tracking during routine operation and to identify the nature and magnitude of the uncertainty due to equipment set up and camera distortions as part of the total uncertainty in a self-consistent manner. We found that uncertainty of the camera frame to global frame rotation matrices produced rotation of the image and uncorrected camera hardware uncertainty produced dilatation or compression of the image twice the magnitude of that seen with camera correction. However, camera resolution remains as an equally important factor limiting the accuracy of photogrammetric tracking that can not be easily reduced numerically. In conclusion, the analysis elucidates how uncertainty propagates to numerical derivatives of the tracking data and prepares the groundwork for future development.