视频影像增强虚拟三维场景的注册与渲染方法研究

发布时间：2018-04-14 10:12

本文选题：虚拟现实 + 增强虚拟　；参考：《武汉大学》2014年博士论文

【摘要】：视频监控系统是当今安全防范系统的重要组成部分,广泛地应用于人们的日常生活,而虚拟三维系统是计算机科学及其相关学科领域发展的重要成果,它的应用也非常广泛。将视频影像融入到虚拟三维系统中以增强虚拟三维场景的真实感,一方面既能够发挥视频影像来源广泛、信息获取方式灵活和实时传输方便的优点,另一方面又能够给虚拟三维系统增加时间维度,将实时、动态的客观世界信息展现在静态的虚拟三维场景中。同时,视频信息融入到虚拟三维场景中还能够克服视频影像分散、无组织的缺点以及虚拟三维场景无动态信息的局限性,从而形成优势互补,从整体上提高二者的应用价值。本文主要是针对当前视频影像增强虚拟三维场景过程中的注册与渲染这两个核心问题展开了相关的研究工作,主要研究内容、研究方法和成果如下： (1)分别分析了视频影像增强虚拟三维场景过程中的注册和渲染这两个核心问题的方法原理：基于视频影像成像模型与虚拟三维场景成像模型的一致性,构建了视频相机内外方位元素与三维渲染引擎相关函数之间的对应关系,可以将后续计算获得的内外方位元素转换为三维渲染函数的相关参数,从而将视频影像准确地注册到虚拟三维场景中；研究分析了视频影像三维渲染的一般性和特殊性,并基于DirectShow流媒体技术设计了视频影像过滤器,能够从不同来源的视频数据中获取实时的纹理图片,并可以利用GPU加速渲染的方法来提高系统的整体效率。 (2)在研究分析视频监控场景的控制条件和先验知识的基础上,探索了视频影像对虚拟三维场景进行注册的方法路线,确定了视频影像内外方位元素分步求解的方法途径,主要包括利用成本较低、操作方便的一维标定法和综合多种特征的后方交会法分别求解视频影像的内外方位元素。 (3)深入地分析了相机一维标定法的误差来源,并在此基础上提出了一种改进的单像一维标定法和一种基于像对基础矩阵的多像一维标定法,这两种方法均能够获得较高精度的标定结果,能够在一定程度上克服一维标定物构造简易、几何约束较差的局限性,适用于视频监控场景中的相机内标定。 (4)基于广义点摄影测量理论提出了一种综合多种特征的后方交会法,该方法能够将水平线、垂直线、线段和水平圆做为控制条件联合常规的控制点构建统一的平差模型从而求解相机的外方位元素,能够在一定程度上摆脱传统后方交会法对控制点的强烈依赖,使基于共线方程的后方交会法具有更广的适用范围,适用于先验知识缺乏、控制点难以布设的视频监控场景。 (5)针对视频影像在虚拟三维场景中渲染所涉及到的数据量与实时性这两个瓶颈问题,着重研究视频影像合理的组织、访问与调度方法来提高视频影像的渲染效率,主要包括：基于多线程技术提出了一种视频数据读取的任务分配和消息通信方法,能够解决虚拟三维系统用户操作与视频数据读取之间的响应冲突；根据近大远小的原则,提出了一种视频影像实时分辨率的计算方法；基于四叉树原理设计了一种视频影像的空间索引；最后,综合上述方法和技术,设计了一种视频数据的缓冲与调度机制,以实现视频影像在虚拟三维场景中的高效渲染。另外,还针对视频相机视角与虚拟三维场景视角不一致的问题,变相地使用了ShadowMap算法来检测其可见性。通过上述工作,本文在很大程度上解决了视频影像在虚拟三维场景中的注册和渲染这两个核心问题,对视频影像融入虚拟三维系统的实际应用有较大的帮助,对增强虚拟技术和视频GIS的发展有一定的贡献价值,对相机标定和视频纹理的相关研究有一定的借鉴意义。
[Abstract]:The video monitoring system is an important part of the security system, widely used in people's daily life, and the virtual 3D system is an important achievement in the development of computer science and related disciplines, its application is very extensive. The video image into a virtual three-dimensional system to enhance the reality of virtual 3D scene. On the one hand can play video sources, flexible information acquisition and real-time transmission convenient, on the other hand can also increase the time dimension, to the virtual 3D system real-time, dynamic and objective world of information displayed in the virtual 3D scene static. At the same time, the video information into the virtual three-dimensional scene can also to overcome the limitations of the video image dispersion, dynamic information organization and disadvantages of virtual 3D scene, so as to form complementary advantages, enhance the overall two The application value of this paper is mainly related to the two core issues of video and image enhancement in the process of virtual 3D scene registration and rendering. The main research contents and methods are as follows:
(1) analyzed video image enhancement principle and method of the two core issues of virtual 3D scene in the process of registration and rendering: consistency of video imaging models and 3D virtual scene imaging model based on the construction of the corresponding relation between the correlation function of video camera interior orientation element and 3D rendering engine, and can be the orientation elements obtained subsequent conversion related parameters for 3D rendering functions, which will be registered to the video image accurately in virtual three-dimensional scene; analysis of the universality and particularity of the video image rendering, and video image filter is designed based on DirectShow streaming media technology, can obtain real-time texture images from video data the source, and can accelerate the rendering method to improve the overall efficiency of the system by using GPU.
(2) in the study of video scenes control conditions and on the basis of prior knowledge of video image registration method on line virtual 3D scene, method of video image and external orientation elements step solution were determined, including the use of low cost, internal and external orientation elements resection method convenient one-dimensional calibration method and comprehensive features of the video images are solved.
(3) the in-depth analysis of the error sources of the camera calibration method is one-dimensional, and put forward an improved single image based on a one-dimensional calibration method and the like on the basis of matrix like one-dimensional calibration method, the two methods can obtain the calibration result with high accuracy, can be overcome to a certain extent one dimensional calibration of simple structure and limitation of the geometric constraint. The scene for video surveillance in camera calibration.
(4) the theory of the generalized point photogrammetry presents a comprehensive resection method based on multi features, this method can be a horizontal line, vertical line, line and circle do to build a unified control point control conditions combined with conventional adjustment model to solve the exterior orientation elements of the camera, can get rid of the traditional method of resection a strong dependence on the control point to a certain extent, the resection method based on the collinear equation has wider application range, applicable to the lack of knowledge of video surveillance, scene control point is difficult to layout.
(5) the two bottleneck data for video image in virtual 3D scene rendering involved with real-time video images, focuses on the reasonable organization, access and scheduling method to improve the video image rendering efficiency, mainly including: multi thread technology presents a video data reading and task allocation message communication method based on virtual 3D system can solve the conflict in response to user operation and read video data between; according to near the small principle, this paper proposes a calculation method of real-time video image resolution; four tree spatial index based on the principle of the design of a video image; finally, the above methods and techniques, design the buffering and scheduling mechanism for video data, in order to achieve efficient rendering of video images in a virtual three-dimensional scene. In addition, the video camera and the virtual three The ShadowMap algorithm is used to detect the visibility of the dimensional scene.
Through the above work, this paper solves the two key problems in virtual 3D scene in the video image registration and rendering to a great extent, the actual application of the video image into the virtual 3D system are of great help, there is a certain contribution to enhance the development of virtual technology and video GIS, have certain reference significance research on camera calibration and video texture.

【学位授予单位】：武汉大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：TP391.41;TN948.6

【参考文献】