MARKERLESS ANALYSIS OF HUMAN GAIT WITH A MULTI-CAMERA SYSTEM

摘要

A model-based approach for markerless analysis of human gait is presented. Images are captured by a calibrated arrangement of cameras. A generic 3D body-model is assumed to be available. The geometrical model parameters (e.g. height and proportions) are defined manually beforehand. To process the image sequences three parameters per motion sequence must be defined manually. Automatic analysis based on a periodic model of human gait then takes place. The motion model parameters are estimated using the method of least squares, whereby all time steps in the sequence are taken into consideration simultaneously. This reduces the number of parameters, and enhances robustness and accuracy.