三维室内场景分析及合成算法研究

发布时间：2018-07-21 11:39

【摘要】：在计算机图形学、计算机视觉等研究领域中,对各种三维场景的建模一直是国内外学者关注的热点,也是极具价值的研究课题。随着计算机仿真和虚拟现实等科学技术的持续快速发展,室内场景的三维建模应用范围也更加广泛,能够应用于火灾救援、人质营救、密室逃脱和考古挖掘等活动中,因此,室内场景建模具有重大的研究意义。目前,物体或场景的建模方法大体分为三种：利用三维软件建模、通过仪器和设备采集信息建模、通过图像或视频建模,通过这些方法建模在时间成本、操作繁杂度和建模效果等方面或多或少存在不足,因此,对于三维室内场景的自动合成成为近几年来场景建模相关领域中的一个极具挑战性的问题。本文分析了场景中物体的几何关系和物体摆放的位置分布,并在此基础上实现了3D室内场景的自动合成方法。论文的主要贡献如下：(1)本文实现了检测与提取场景中物体之间的相交、接近、支撑三种空间关系的算法,同时实现了对场景中支撑平面的提取工作,并构建了物体支撑的共生关系。对场景的几何分析为优化场景合成算法提供了理论依据。(2)本文采用了GMM拟合物体在支撑面上的摆放位置,得到了位置分布模型。首先建立一个文件,将统计得到的物件在支撑面上的归一化位置存放其中,坐标个数代表了物件在场景库中出现的次数。然后用高斯混合模型拟合该文件内容,得到位置分布模型。(3)本文提出了一种三维室内场景的自动合成方法,由用户给定的初始室内场景和要添加到场景中的小物件描述,借助素材库选择小物件,利用场景合成算法实现这些小物件的自动摆放,展示了简单场景不断丰富的演化过程。该方法即体现了用户的设计需求,又避免了繁琐的人工操作。
[Abstract]:In the fields of computer graphics and computer vision, modeling of various 3D scenes has been a hot topic and a valuable research topic for scholars at home and abroad. With the rapid development of science and technology, such as computer simulation and virtual reality, 3D modeling of indoor scene is more widely used in fire rescue, hostage rescue, secret room escape and archaeological excavation. Indoor scene modeling is of great significance. At present, there are three methods for modeling objects or scenes: modeling with 3D software, collecting information through instruments and devices, modeling by image or video, modeling at time cost by these methods. The operational complexity and modeling effect are more or less inadequate, so the automatic synthesis of 3D indoor scene has become a challenging problem in the field of scene modeling in recent years. In this paper, the geometric relationship of objects and the position distribution of objects in the scene are analyzed, and on this basis, the automatic synthesis method of 3D indoor scene is realized. The main contributions of this paper are as follows: (1) this paper realizes the algorithms of detecting and extracting the intersection, proximity and support of three kinds of spatial relations between objects in the scene, and at the same time realizes the extraction of the support plane in the scene. The symbiotic relationship of object support is constructed. The geometric analysis of the scene provides a theoretical basis for optimizing the scene composition algorithm. (2) in this paper, the GMM is used to fit the position of the object on the support surface, and the location distribution model is obtained. Firstly, a file is established to store the statistical objects in the normalized position on the support surface. The number of coordinates represents the number of times the objects appear in the scene library. Then the Gao Si hybrid model is used to fit the contents of the file and the location distribution model is obtained. (3) in this paper, an automatic synthesis method of 3D indoor scene is proposed. The initial indoor scene and the description of small objects to be added to the scene are given by the user. With the help of the material library to select small objects, the scene composition algorithm is used to realize the automatic placement of these small objects, which shows the rich evolution process of simple scenes. This method not only reflects the design requirements of users, but also avoids tedious manual operation.
【学位授予单位】：北京交通大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP391.41

【相似文献】