一种改进算法的光学运动捕捉系统

对光学运动捕捉系统中关键技术进行了论述，为了得到无需后期手工标注的三维数据，需对二维标记点预测和跟踪，提出了利用先验数据准确跟踪标记点的算法，采用非线性和线性结合的方法标定摄像机，计算出摄像机的外部参数和内部参数，三维重建时利用多个摄像机融合重建，提高了重建的准确率，并利用对摄像机分组的方法，降低了运算的复杂度，实验结果证明，本系统能够得到较为精确的人体运动数据。

Optical Motion Capture System with Improved Algorithms

The key techniques of optical motion capture system were discussed. In order to get the 3-dimensional data dispensing with late manual labeling, we predicted and tracked the 2-dimensional markers. Therefore, an algorithm using the prior data to predict the markers was proposed. The non-linear method combined with the linear one was used for calibrating the cameras for computing their extrinsic and intrinsic parameters. Using a multi-camera to recover the 3-dimensional data converged- ly,the accuracy of the reconstruction was improved. The computational complexity was reduced by grouping the cameras. The results show that our method can get the accurate human motion data.