Calculation of the residual error in three-dimensional videography using stationary, panning and mobile video capturing techniques

Abstract

Significant technological advancements have been made in the study of human motion over the past few decades. There are three known video capturing techniques for collecting three-dimensional kinematic data; stationary, panning and mobile camera set-ups. The panning and mobile camera techniques allow for the cameras to follow a subject throughout the entire length of the skill while maintain a close-range view of the subject, something the stationary camera set-up does not allow. Accuracy of the panning and mobile camera techniques has not yet been well documented. The main objective of this study is to determine root mean square (RMS) errors of three-dimensional coordinates in a 5.5-meter filming volume using stationary, panning and mobile camera techniques using three specific testing conditions. Although the three camera techniques showed acceptable residual errors throughout the three testing conditions, the mobile technique produced the lowest RMS error for the two out of the three test condition. For the panning and mobile camera set-ups, camera-to-subject distance is much smaller than the static camera technique for a 5.5-meter field. Lamontagne et al reported that residual errors could be minimized when camera-to-subject distance is reduced. The mobile technique produced acceptable accuracy in video analysis (Lafontaine et al., 2000). This technique did reduce the residual errors, of positional coordinates of markers and the vectors length measurements; by 55% and 44% in the gait testing condition respectively from errors obtained using the stationary camera technique. This study allowed for a comparison of three video capturing techniques abilities to collect accurate data over a large filming area. The results show that, although all three camera set-ups produce low residual errors, the mobile camera set-up offers the best accuracy while maintaining a small camera-to-subject distance, allowing for close range studies to be performed successfully.