Design And Implementation: A Method Of Motion Data Acquirement Based On Stereo Measurement

Motion data acquirement is the key technology of computer aided sports training system. To provide more expeditious and accurate data for motion simulation, we propose a method of motion data acquirement, based on stereo measurement. This method includes two parts, motion tracking and motion modeling.In the process of human motion tracking, first of all, two or more high speed cameras are used to synchronously film a athlete's motion process, and the motion videos are collected; secondly, using human motion constraints and the mean-shift marker-less tracking algorithm, the pixel coordinates of the human skeletons in each image can be tracked and captured. Experimental result shows that the improved mean-shift algorithm can track the monitoring points precisely. Even when the monitoring point is blocked out, the position of the monitoring point can be speculated on according to the motion constraints.In the process of human motion modeling, firstly, we use the Directly Linear Transaction (DLT) algorithm to calibrate the cameras, and then use the Levenberg-Marquart algorithm to optimize the projection matrix; secondly, by using the stereo vision theory, the 3D coordinates of each skeleton of the athlete are calculated to model the human motion; finally, the human motion data are optimized by solving the energy function of human body constraints. Experimental result shows that using these motion data, the human motion can be simulated well and the motion indexes can be calculated accurately.The thesis's work points out that this motion data acquirement method is of some robustness and adaptability. This method can largely reduce the human-computer interaction, reduce user's workload and improve work efficiency.