Research And System Development On Near-infrared Optical Motion Capture Technology

Recently, motion capture technology is developing quickly and has been widely used in education, medical treatment and entertainment. Many companies developed mature motion capture system one after another. Most of these systems can be satisfied with the requirement of motion capture but they are expensive. To many organizations it is impossible to have such motion capture system.This paper is mainly about how to construct a motion capture system based on near-infrared cameras. We discuss some key procedure and technology of the motion capture (MoCap) system, and propose a MoCap framework. We implement a MoCap system according to the framework with lower cost than others.For the calibration step we use a muti-camera self calibration algorithm based on one marker. This algorithm uses the pixel coordinate set of the marker in different camera projection plane on different time point as input and calculates the interal parameters and external parameters of each camera using some approaches of computer vision.Use the projection matrix output by calibration step to do motion reconstruction. To solve the problem of matching the markers among projection plane of different cameras, we use epipolar constraint together with reprojection error evaluation. Another problem is matching the markers between two consecutive frames. We develop an algorithm based on iterative comparison to get the solution of this problem.We also do some research work on motion data processing mainly about filling missing markers. By comparing different filling methods we find a reasonable strategy to do this job and implement these methods in the MoCap system.The experiment shows that our motion capture system captures human motion data robustly in real-time.