Research On Key Problems In Large Real-time Interaction Augmented Reality System

Augmented Reality (AR) appears with the rapid development of virtual reality and expands the virtual reality. And it enhances the real world by superimposing the information computer generates on top of it. Past decades have witnessed its great success in calibration technology and display technology. Recently, human computer interaction (HCI) in AR has attracted more and more people's attention. Particularly, tangible AR becomes the new research focus.The essence of TUI is that the people's interaction with objects and environment in real life is mapped into the interaction with virtual objects in virtual world. So users can utilize habit and experience of their lives in HCI directly. Grasping and throwing are used frequently in daily life, so taking grasping and throwing virtual objects as the interaction of large AR system is bound to increase interactive naturalness and convenience. But it is a complex process that includes registration, tracing, mutual occlusion, embedded 3D stereoscopic imaging, image fusion, controlling magnetic force trackers'mutual interference, etc. This paper studies in three areas: the reliability of tracking, naturalness of interaction and the reality of response, and achieves the following results:1. In large intaction augmented reality system magnetic force transmitters are placed too near with one another, which will make mutual interference, and virtual object in helmet vibrates obviously. This paper presents a controlling interference strategy based on particle filter, the Kalman filter and the improved intrinsic 3D discreet Curve smoothing algorithm for tracing head and hand moving. To the interference of tracing hand, a particle filter is adopted. To the interference of tracing head, while head holds still, the Kalman filter is used. While it movs slowly, the improved intrinsic 3D discreet Curve smoothing algorithm is used. This improved algorithm is combined with approaching-real-time model and bilateral filtering, and it realizes the using of static de-nosing algorithm of graphics in the real-time filtering to 6DOF dynamic trajectory. While it moves quickly, filter stops. Furthermore, the critical damping is applied to avoiding jump while entering the quick state.2. By using monocular optical see-through HMD, the augmented reality system can make an optical illusion for the users, where the virtual object is in front of real objects. So the users can't make sure where they should grasp the virtual object. This paper presents a grasping recognition method based on visual awareness. By the custom of person's visual awareness, the collision detection based on visual awareness is realized from the robotics and geometrical optics theory. Three knowledge rules about grasping are gained by analyzing grasping action, and then the rules are fused with the fuzzy theory in the evidence deduction.3. For the limitation of existing hardware and technical conditions, space between virtual object and fingers often emerges after recognition, which may be perceived by user if they haven't been occluded. So the grasp is not natural enough. This paper presents a new grasp's natural confluence method. Making use of 3D graphics'theory the occlusion between virtual object and fingers can be judged while head or hand is moving. In the light of this judgment and grey theory, the hand's movement can be predicted. Moreover, according to the prediction's result and the principle of max speed difference, the virtual object's velocity can be adjusted adaptively to realize natural conflation.4. The response reality of virtual projectile is an important aspect of interaction reality, while means the object's posture and trajectory should be decided by the throwing. This paper presents a throwing recognition method based on magnetic force tracking and projectile's real-time rotation physical model. And it dooms to find out the character of throwing and the reasons of rotation and the approach to rotation by using the theories of rigid body kinematics, Aerodynamics and the velocity changing law of throwing. In addition, the relation between projectile motion and throwing can be established. By this way, projectile has realistic posture.Experiments show that the mutual interference is reduced significantly by using this controlling interference method, especially the improved intrinsic 3D discreet Curve smoothing algorithm, which can control larger interference well while moving slowly. The grasping recognition method based on visual awareness has the great capability of processing uncertain information, and the users can grasp virtual objects without hint. The natural confluence method has high reliability which can limit the prediction error, and high flexibility and real-time which can adjust adaptively according to hand's moving. The throwing recognition method based on magnetic force tracking and the projectile's real-time rotation physical model whose computational performance is accurate and efficient. The users can throw virtual objects wearing only one magnetic force receiver, and the projectile's posture has closed relation to throwing, which gives them experiences of throwing real object.