增强现实中深度一致性问题的研究

发布时间：2018-02-16 11:12

本文关键词： 增强现实深度一致性虚实融合双目立体匹配　出处：《沈阳工业大学》2017年硕士论文　论文类型：学位论文

【摘要】：增强现实是目前科技信息技术应用领域中热门的话题,其技术是指将由计算机生成的虚拟信息(如图形图像、音频等)与现实场景中的真实物体融合一体,并将融合后场景展现给用户,增强用户对现实环境的认识与感受。增强现实技术涉及到显示技术、图像识别技术、图形学、机器视觉、虚实融合遮挡等多门学科,在各个领域(如生活娱乐、医疗、航天等)都被广泛应用。随着增强现实应用领域的需求与相关技术的不断研究发展,用户关注点转移到计算机生成的虚拟信息与真实场景的融合深度一致性问题上。在目前已有的增强现实系统中融合深度处理仅是将虚拟信息放置在二维场景图像中,从而合成虚实物体融合场景,但这将导致虚实物体的深度关系无法改变,缺少空间的深度效果,最终使用户不能感知融合后的真实感。对于上述虚实物体融合深度问题,本文经过分析总结现有处理虚实物体融合深度关系的方法,并结合图形学、计算机双目视觉等多门学科的理论,提出了一种基于双目摄像机的深度计算方法来解决增强现实中深度一致性问题。该方法首先采用基于棋盘模板对双目立体摄像机进行双目标定并获取摄像机的内外参数;其次为了实时获取视差图,引入动态规划优化方法的半全局立体匹配法SGBM(Semi-Global Block Matching)设计一种实时获取视差图的方法;然后利用反投影三角相似测量原理求解出三维场景物体的深度信息;最后经过对比虚拟物体与真实场景中的物体上像素点Z轴坐标值(深度信息)大小并判断虚拟物体是否在真实物体轮廓内来对场景进行渲染,进而解决了深度一致性的问题。通过实验验证,实时获取的真实场景的三维深度信息的方法具有光照不变性、准确性。
[Abstract]:Augmented reality is a hot topic in the field of application of science and technology information technology, which refers to the integration of virtual information generated by computer (such as graphics, images, audio, etc.) with real objects in real scenes. And the fusion scene is presented to the user to enhance the user's understanding and feeling of the real environment. The augmented reality technology involves many subjects such as display technology, image recognition technology, graphics, machine vision, virtual reality fusion occlusion and so on. It has been widely used in many fields (such as entertainment, medicine, spaceflight, etc.). The focus of user is on the consistency of fusion depth between computer generated virtual information and real scene. In the existing augmented reality system, the fusion depth processing only places the virtual information in two-dimensional scene image. Thus, the fusion scene of virtual and real objects can be synthesized, but this will lead to the immutable depth relationship of virtual and real objects, and the lack of depth effect of space, which ultimately makes the user unable to perceive the reality after fusion. For the problem of the fusion depth of virtual and real objects mentioned above, This paper analyzes and summarizes the existing methods of dealing with the relationship between the fusion depth of virtual and real objects, and combines with the theories of graphics, computer binocular vision and so on. In this paper, a depth calculation method based on binocular camera is proposed to solve the problem of depth consistency in augmented reality. Firstly, the binocular stereo camera is determined based on the board template and the internal and external parameters of the camera are obtained. Secondly, in order to obtain parallax graph in real time, the semi-global stereo matching method (SGBM(Semi-Global Block matching), which is a dynamic programming optimization method, is introduced to design a real-time parallax image acquisition method. Then the depth information of 3D scene object is solved by using the triangulation similarity measurement principle of backprojection. Finally, by comparing the size of the Z axis coordinate (depth information) of pixels on the virtual object and the object in the real scene, and judging whether the virtual object is in the contour of the real object to render the scene. Finally, the problem of depth consistency is solved. It is proved by experiments that the method of acquiring the 3D depth information of real scene in real time has the characteristics of illumination invariance and accuracy.
【学位授予单位】：沈阳工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.9

【参考文献】