Estimation of six degree of freedom rigid body segment motion from two dimensional image data

The human body is often modelled as a set of linked-segments moving in space. Each segment may be considered rigid, and hence, will move with 6 · degrees of freedom (df). Conventional photogrammetric kinematic methods estimate Rigid Body (RB) motion through knowledge of the locations of 3 or more non-collinear points on the RB. These points are measured in two positions, one control, the other transformed. Methods of Three Dimensional Reconstruction (3DR) are commonly used to obtain these measures. Within 3DR, each spatial marker requires coincident observation from at least two images. If two or more observations are unavailable, 3DR cannot proceed, thus potentially valid marker observations from a single image cannot be used to assist in 6 df measurement of an RB. Insufficient 3DR may result in regions of incomplete 6 df data during time-series analyses of motion. This paper presents a method, TRACK26, which estimates 6 df directly from all available two dimensional image data. Use of this method with 4 or more non-collinear markers on an RB can help to eliminate regions of missing 6 df kinematic data. An experiment with a uniformly revolving eleven marker RB shows TRACK26 to be more accurate and more successful, in practice, than a typical conventional RB tracking method. Where success is judged as being the ability to provide 6 df estimates, the conventional method returned a success rate of 46.6% compared to 100% for TRACK26. If one considers the success rate of determining 3D points alone, 3DR was found to be inferior to either of the RB tracking methods. It is also noted that TRACK26 can be used to provide 6 df RB motion estimates with the use of a series of single images. Given these advantages it is concluded that TRACK26 can provide more accurate and complete data for use in linked-segment analyses of human motion.