大场景三维重建系统研究
文献类型：学位论文
| 作者 | 张峰 |
| 学位类别 | 工学博士 |
| 答辩日期 | 2009-05-28 |
| 授予单位 | 中国科学院研究生院 |
| 授予地点 | 中国科学院自动化研究所 |
| 导师 | 胡占义 |
| 关键词 | 三维重建 系统标定 直线匹配 自标定 三维配准 3D reconstruction system calibration line matching self-calibration 3D registration |
| 其他题名 | Study on Vision Systems for Large Scene Reconstruction |
| 学位专业 | 模式识别与智能系统 |
| 中文摘要 | 视觉系统的研究是当前三维计算机视觉研究的发展趋势，本文以古代建筑三维快速建模及犯罪现场复原为应用背景，在系统标定，结构化室内场景重建及大范围室外场景自动重建三个方面进行了比较深入系统的研究，主要工作有： 1．针对自行研制的基于旋转平台的室内重建系统，给出了两种基于坐标变换的系统标定方法。研究发现该系统的标定实际上是一种特殊的手眼标定问题，通过合理地选取坐标系，可简化该问题的求解。本文给出的第二种基于坐标变换的方法，标定过程更为灵活方便，而且鲁棒性好，标定精度高，基本能满足实际应用的需求。 2．给出了一种全自动的室内大范围场景的重建系统。由于室内场景具有结构化的特点，如人们习惯的平行、垂直、共线、共面等，在基于图像的室内场景自动重建中，即使一些小的误差也可以导致明显的视觉差异。针对犯罪现场复原的具体需求，给出了一种基于图像的室内场景自动重建系统。该系统有如下特点：重建过程为全自动，不需要任何人机交互；直线特征的自动匹配与重建考虑了场景的深度与结构信息，匹配的正确率及直线重建效果得到了显著提高；重建过程的整体优化中，融合了特征点与特征直线。 3．针对古代建筑快速建模的具体需求，本文给出了一种基于图像序列的室外远距离场景的自动重建系统。主要内容有：讨论了相机模型的选取问题，给出了一种适用于大范围场景的透视相机模型；给出了一种仿射相机模型的鲁棒的自标定方法；利用分解法及分块的捆绑调整技术很好解决了多视角下重建结果的融合问题。 |
| 英文摘要 | The study on vision systems is one of the hot topics in computer vision. Aimed at the 3D recovery of ancient Chinese architectures and that of crime-scenes, the thesis is focused on some of the fundamental problems of vision system, including calibration of a 3D reconstruction system with rotating platform, 3D reconstruction of structured indoor scene and automatic 3D reconstruction of large-scale outdoor scene. The main works are summarized as follows: 1. Two different system calibration methods for the rotating platform are proposed. In fact the system calibration is found to be a special case of the hand-eye calibration. By properly choosing the coordinate system, the calibration can be largely simplified. In addition, experiments show our second system calibration method based on coordinate transformation is not only flexible and robust, but it is also of good accuracy. 2. An automatic image based 3D reconstruction system for indoor scene is presented. Due to the structured nature of indoor scenes, such as the perpendicularity, parallelism, colinearity, coplanarity etc, accustomed by human being in their daily life, usually a small reconstruction error of indoor scene could bring up dramatic visual effects. In this work, an automatic image-based crime-scene reconstruction system is obtained, which includes such units as image capturing platform calibration, image point and line matching and reconstruction as well as the fusion of partial results under multiple views. The main characteristics of our system are: fully automatic, no human interaction is needed; both structural and depth information is used in line matching and reconstruction to substantially increase the reconstruction quality; in the bundle adjustment, both lines and points are optimized simultaneously. Extensive experiments on real scene validate our system. 3. Aimed at 3D modelling of ancient Chinese architecture, an automatic 3D reconstruction system of distant outdoor scenes from images is presented. The system has the following advantages: (1) A more suitable perspective camera model for large-scale outdoor scene is proposed. (2)A robust self-calibration method based on affine camera model is introduced. (3)A combination of the factorization method and the bloc-based bundle adjustment technique is adopted for the efficient fusion of partial results under multiple views. |
| 语种 | 中文 |
| 其他标识符 | 200618014628043 |
| 源URL | [http://ir.ia.ac.cn/handle/173211/6176]  |
| 专题 | 毕业生_博士学位论文 |
| 推荐引用方式 GB/T 7714 | 张峰. 大场景三维重建系统研究[D]. 中国科学院自动化研究所. 中国科学院研究生院. 2009. |
入库方式： OAI收割
来源：自动化研究所
浏览0
下载0
收藏0
其他版本
除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。