The Motion Capture Data Aided Training In Applied Research

Motion Capture is to record the motion track in three dimensions by sensor devices, to transform it to motion data, and then to drive the virtual human to move based on these data.In recent years, there are a lot of works done on the Motion Capture outside nation and many systems on Motion Capture have been provided to use. But all of the systems are so expensive that few institutes or companies in our country can afford. A few institutes in our country have been engaging in the business for several yeas, however, the images collected by the system made by them are often planar and have no three-dimensional motion parameter because the images are captured by cameras and have to be disposed on sensibility experience. As a result, it is very necessary to design a 3D Motion Capture system by ourselves to fulfill the internal demand.Based on "TRI-EYE" which is a device collecting spatial information made by CFR, this dissertation aims at real-time and reality, reaches a new precision level in foundation of guaranteeing it's real-time, and uses the device to assist training, and create an applied training system based on motion capture.The significance of advancing the reality and real time of motion-capture is summarized and the development actuality of the motion-capture systems is depicted in the first chapter. The factors that may demolish the consistency of motion capture and some problems needed further research are argued in this chapter. Also, tactics for implementing the assist-training system are introduced here.By analyzing scattered data fitting theory, the second chapter compares several data-fitting methods, propose a layered and repetitious data-fitting algorithm based on searching, explain the algorithm in detail and present the experimental results.By analyzing calibration methods to vidicons at home and abroad, the third chapter puts forward an outside calibration algorithm used in this system. The algorithm is used to ascertain the three-dimensional position and direction in the world reference frame of the spatial information collection device and in order to get the world coordinate of the collecting data. The chapter explains the algorithm in detail and presents the experimental results.After studying various models of virtual humans, an octahedron model is selected in this system and the octahedron model is drove to display in a real time in the forth chapter. And after analyzing the enging——Cal3D, this chapter puts forward a tactic to create Csm file. In the endof the chapter, a data transform method from Csm to files of Cal3D is narrated in order to show the trainee's action through Cal3D.The fifth chapter excogitates some methods to assist golf training based on analyzing other assist-training systems with motion-capture devices. These methods can be used to correct the incorrect gestures of the trainee.Finally, the 6th chapter argued the problems that still exist home, and gives some advices for later research in detail.