Posture Kinematics Reconstruction and Body Model Creation

摘要

The application of the proposed method allows to analyse images recorded during orthostatic posture trials. The possibility to put in evidence mechanisms that govern postural control has been demonstrated evaluating the values of hip rotation over time. The algorithm, furthermore, produces a human body model. In particular, the method is able to estimate the trajectory of the typical kinematic variable, the Centre of Mass, which gives valuable informations to clinicians. In order to quantify the movement ability of the patient, the exploitation of image recording techniques inside movement analysis laboratories has become necessary in clinical practice: their spreading out and thus their routinely use is, however, prevented by the non-soberness in economical and timing terms, caused by the use of "ad hoc" motion analysis systems, and the setting up of external devices that have to be applied to the patient. In this context, the use of commercially available video systems, together with the development of techniques granting the reconstruction of the kinematics in absence of markers, reduces costs and times, and therefore favours their diffusion in clinical environments. However, novel methods for image processing must be used in order to determine the position of body segments during orthostatic exercises with the required accuracy. The method proposed in [3] has been used as the kernel for the motion estimation process into video sequences. It has been tested both in simulated motions, and in real video sequences, and applied to the tracking of body points during orthostatic postural tests, and has been demonstrated able to reliably represent the kinematics of the stabilometric fluctuations in the sagittal direction; in particular, the method is capable to discriminate between mechanisms that govern the human plant in postural tests, and then to give useful information for the clinician. As an extension of the technique to other Human Motion Analysis applications, the technological development facilitates the extension of the limits for frame rate acquisition: during the recordings of biomechanical tasks, if the motion process is temporally oversampled, the disparity between time-adjacent frames will be smaller, and the approximation driven by simple translational motion ensured if the dynamical update of the reference is used. Correspondingly, the motion estimation process must be more accurate, as the relative motion of the selected elements is minor; in this context the interpolation process to subpixel resolution is the key for every estimation process in human movement analysis.