Based On The Augmented Reality Research And Implementation Of Real-time Interactive Technologies

Humans are bringed into wonderful virtual world by computer science, so we cansimulate real world activities in the virtual environment. People have strongly desire toexplore new areas and no longer satisfy immerse in the virtual world. The developmentof virtual reality technology enters into a more important field–augmented reality.Compared to VR, AR bringing people from the real world completely into the virtualworld, augmented reality take the virtual information into the real world in order thatpeople can perceive the existence of virtual objects in the real world. With the rapiddevelopment of computer technology and multimedia technology, people urgent tointeract with virtual information by some natural way, such as posture, facialexpressions, voice interaction and so on.This thesis studies virtual and reality interaction technology based on naturalpicture identification and semantic analysis of human3D skeletal reconstruction andidentification. Firstly, it introduces the calibration technique and algorithm based oncamera model. Secondly, it focuses on3D recognition method of natural feature point,extracting and matching natural feature points by SURF algorithm and using RANSACalgorithm to remove the mismatch feature. L-K optical flow algorithm is combined withimage pyramid used to track feature points in order to improve the augmented realitysystem efficiency. Taking into account the tracking instability, this thesis presents amethod to reduce unrelated number of feature points by last time of point locationinformation when tracking fail.Users hope wanderfully experience both on visual and natural interaction withvirtual object. So the thesis discusses hand interaction based on semantic analysis ofhuman3D skeletal reconstruction and recogintion. It analyzes the method of removingthe background and extracting the Body shape on depth map. And then it introduces akey technology which marks body and get identify of3D bone joints information bymachine learning. Finally, we get the joint position information by OpenNI combinedwith PrimeSense Series’ camera. And we use finite state machine principle to define thegesture and realize interaction bewteen the smart gestures and virtual objects. The research is applied to the actual project development and realized sub-moduleof AR system of the JinSha museum with OGRE. We use gesture real-time interactivewith virtual artifacts which recognition by natural picture to enhance the experience ofusers.