Research On Tracking And Reconstruction Of 3D Human Gait Based On Optical Markers

Human gait analysis plays an important role in rehabilitation.The key issue in gait analysis is gait motion capture. Although commercial motion capture systems such as VICON prevail in many countries, their use in China are far from popularity.This situation is partly attributable to their high price and complexity.This thesis aims to develop a low-cost,easy-to-use 3D human gait capture system for rehabilitation medicine. The main contributions of this paper are as follows:We present a scheme for maker-based 3D human gait capture using multiple synchronized cameras. The approach consists of five phases, that are camera calibration, 2D marker detection, 2D marker tracking, 3D marker reconstruction and motion data estimation.We propose to utilise a multi-camera self calibration method to estimate camera parameters. The method simultaneously calcuates the internal and external parameters with considerable accuracy.The re-project error is less than one pixel.We devise a self-correcting tracking approach for 2D marker tracking in image sequences. The approach predicts maker positions with Kalman filter and than the position is corrected using a smoothness constraint. The corrected positons are further verified by epipolar geometry and error correspondences are rejected.We propose an accurate and robust joint parameter estimation method based on function analysis. The method includes a data preprocessing step, which is based on rigidity modification method and can improve the quality of the current joint parameter estimation methods. Joints come from this initialization are more accurate and robust than from others.We develope a prototype system for 3d huamn gait capture. The system implements the aforementioned methods and is integrated in a 3D human gait analysis system. We have ran several experiments on this system and the results exhibited the effectiveness of the proposed methods.