The 3D Simulation And Analysis System Of Human Motion

The famous Olympic Games will be held in 2008 in Beijing, the Chinese capital. To hold a most excellent Olympic Games of science and technology, numeric Olympic Games, a new concept, was proposed. It becomes a very important thing to supervise and train the athletes scientifically, which needs a good and practical simulation and analysis system. At the present time, there are some professional softwares. For example, Ariel Performance Analysis System (APAS), developed by Prof. Ariel and his team, are applied widely in sport and healing fields. Engine- Measure And Analysis System on Sport Images (Engine-System), developed by engine-human information graduate school, are used by many groups. However, these softwares have some problems:(1) based on foreign models but instead of Chinese models to calculate data, such as the centroids of human parts, the human barycenter;(2) only render some simple line-frame pictures, not realistic human models, short of verisimilitude;(3) the recoveried poses of models can not be modified;(4) the frames are not convenient to install or use;(5) the recognition of human keys totally depends on handwork, which needs a lot of job to be done;(6) the prices are very expensiveConsidering the Chinese situation, I developed the 3D simulation and analysis system which is suitable for Chinese models. Here is some features of the system:(1) designed a frame which is easy to install and use;(2) based on Chinese models the first time in China to calculate the sport data;(3) recognise the frame keys and the human keys by computer automatically and users;(4) apply 3D human models to render intuitive scenes;(5) the model poses can be modified by users.The system uses two common cameras to import sport images from two different points of view and generates intuitive model's sport frames. Here is the main job:The human key points in the images should be recognised at first. The work to recognize the key points is so heavy that it's better to do it by computer automatically.My classmate is doing the research now. In my research, the computer can recognise quickly the key points of the images which are marked in advance. If the key points are not marked, I made a human line frame of reference to help users to drag the key points to the right position. Other congeneric softwares always recognise the key points totally by users, for instance, users have to point out the key points in a given order in APAS, and can not modify the worked images. While, in Engine-System, users have to point out one key point in every image, and then another point,until all the key points are pointed out. A user who used the software told us, it would spend many hours to point out all the key points of motion images of standing long jump. Therefore, to help users point out the key points expediently, I made a default human line frame whose key points can be dragged at will by users. Users can modify every worked image, save the file anytime and go on their job next time. I use the same method to recognise the key points of the frame.Based on the recognized 2D coordinates of the frame and the Direct Linear Transform (DLT) arithmetic, we can figure out the parameters of every camera. Then we can figure out the 3D coordinates of the human key points based on their 2D coordinates, the camera's parameters and DLT arithmetic. The 3D human line frame picture can be shown by linking the 3D points correspondingly, which are the final result in other softwares. To render a intuitive scene and to be convenient to modify the model's poses by users, I did a deeper research and made an entity model and a realistic framework model. We can figure out the angles of every human part and their rotation axes based on the 3D coordinates of the human key points, then act them on the 3D models to generate the right poses.To make it easy to modify the model's poses, I made the key-chain structure and t...