Font Size: a A A

The Research On Key Technology Of Visual Feedback In Augmented Reality

Posted on:2016-03-22Degree:MasterType:Thesis
Country:ChinaCandidate:Y R YanFull Text:PDF
GTID:2308330479495431Subject:Computer application technology
Abstract/Summary:PDF Full Text Request
Augmented Reality(AR) is a branch of Virtual Reality, which mixes together the real environment and virtual object, virtual scene or prompt message generated by the computer by using modern display technology, human-computer interaction technology and computer graphics and multimedia technology. Users could find that the virtual information generated by the computer to be a part of the real world around them through AR technique. The key technology of AR mainly includes tracking and registering, virtual-real fusion, lighting, displaying techniques and interaction techniques. All these techniques were aimed at assuring the virtual information added in real world to be accurate so that users could find the system more real and immersive. The most acceptable definition of AR is: 3D registration, virtual-real fusion, real-time interaction.Considering the definition of AR, the three aspects were researched respectively in this paper. Proceed from the AR system, some in-depth investigations and researches were done in relevant three aspects: 3D registration based on natural features, virtual-real fusion consistency and real-time interaction based on 3D model. The main contributions of this paper are as follows:1. A mark-less register method based on non-linear scale space was proposed. The existing register method would fail to register when the motion blur occurs, which was caused by the relative displacement between the camera and the target. To address this problem, the method proposed here could also find the right feature point in motion-blur image for further register. Firstly, a tracking method based on compressed sensing was used to track the register target. In this way, the range of feature detecting would be diminished. Secondly, KAZE was adopted to extract the feature points over the range. Lastly, low-dimension feature descriptors were generated by using principal component analysis(PCA) method to increase the running speed. By matching the feature descriptor and pattern descriptor of each frame, calculating the homography matrix by affine reprojection and registering on the real scene in each frame, the tracking and registering were realized.2. An analogy procedure of virtual object’s motion consistency based on AR system was proposed. For the sake of meeting the requirement of seamless fusion between real environment and virtual object, we get several steps to make the virtual and real objects to achieve the same effect. First of all, Radon conversion and Canny operator were both used to get the motion-blur parameter estimation. Generally, the motion-blur parameter means the point-spread function(PSF) and it includes motion direction and measure. After that, blur rendering would be made to the virtual object combining PSF. Open GL accumulate buffer was used as a tool to do rendering. The method we proposed can effectively increase the consistency of fusion in motion-blur simulation and enhance the immersion sense of users.3. The virtual-real interaction technique based on 3D model was proposed. Firstly,.x model file was imported into AR system by using Direct3 D. This kind of model includes grid skin information, skeleton level information and animate information. Then, the user’s actions would be captured by the Kinect camera and skeleton tracking would be done. Kinect can track 20 articulations to get user’s real-time action feedback. By making the Kinect skeleton tracking corresponds to.x model’s skeleton controlling, users can control the model in real time. The method proposed in this paper gave users more natural interactive experience and provided new method for 3D animation production in AR system.Researches were done in this paper around the visual feedback techniques in AR, including mark-less registration in move condition, fusion consistency based on mark-less registration and virtual-real interaction based on 3D model. It is shown that the AR system designed in this paper is innovative and effective in registering, fusion and interaction. And the researches we discussed here will provide certain reference for the development and application of augmented reality.
Keywords/Search Tags:Augmented Reality, Mark-less Register, Non-linear Diffusion, Motion Blur Simulation, Skeleton Tracking
PDF Full Text Request
Related items