无人机图像的三维重建方法研究
文献类型：学位论文
| 作者 | 郭复胜 |
| 学位类别 | 工学博士 |
| 答辩日期 | 2013-05-28 |
| 授予单位 | 中国科学院大学 |
| 授予地点 | 中国科学院自动化研究所 |
| 导师 | 胡占义 |
| 关键词 | 无人机图像 三维重建 辅助信息 分组重建 批处理重建 真正射影像 重建原型系统 UAV images 3D reconstruction auxiliary information view clustering batch reconstruction true orthoimage reconstruction prototype system |
| 其他题名 | 3D Reconstruciotn from UAV images |
| 学位专业 | 模式识别与智能系统 |
| 中文摘要 | 本文旨在将近年来在视频、图像领域获得巨大成功的三维重建技术应用到无人机图像处理领域,对无人机图像进行全自动重建相关的应用进行研究和探索,可实现高效、鲁棒、批量化无人机图像处理应用.本文的主要工作包括以下几个方面: (1) 基于辅助信息的分组融合重建方法:将图像先分组重建,然后融合的方法,是解决大场景三维重建中规模问题的最有效的途径.无任何先验信息下的图像分组,不仅计算量大,而且很难取得有效的分组结果.目前很多相机均带有GPS、指南针等辅助装置,尽管这些辅助装置提供的信息精度不高,但合理利用这些粗略的辅助信息,可望有效简化大场景三维重建中的图像分组问题.本文对基于这些粗略辅助信息的图像分组和重建问题进行了探讨,给出了一种基于视图重叠度的图像分组方法,并通过大量实验验证了该方法的有效性. (2) 基于辅助信息的批处理方法:首先分析了经典增量式三维重建方法Bundler在无人机图像处理中存在的问题,在分析无人机图像的辅助信息的特点的基础上,提出了一种基于辅助信息的批处理三维重建方法.多组无人机图像三维重建实验表明：本文提出的方法在算法鲁棒性、三维重建效率与精度等方面都具有良好的表现. (3) 大比例尺真正射影像生成:将计算机视觉中获得巨大成功的多视三维重建技术应用到对无人机影像处理中,给出了一种基于运动恢复结构重建算法和多视图立体视觉算法全自动生成大比例真正射影像的方法.本文首先分析了无人机图像PMVS重建点云的特点,给出一种由基于面片多视图立体视觉稠密点生成数字表面模型的方法,然后详细介绍了包括正射影像图像坐标映射模型、可见性计算、基于马尔科夫随机场能量优化的面片选择和匀光处理等真正射影像生成的关键步骤.实验结果以及与商业软件的比较表明：本文给出的方法在野外地形和城市区域均能获取有效的真正射影像结果. (4) 三维重建原型系统的研发:研发了一套具有自主知识产权的三维重建原型系统.该系统集成了多种可选则的算法,可以鲁棒、高效地处理包括地面获取图像和无人机图像的三维重建.本文对系统的构建、重建结果、精度测试结果以及系统性能等方面进行了简要介绍.经过包括地面图像和无人机图像实际的应用测试,表明该原型系统在一定程度上能满足实际重建应用的需求. |
| 英文摘要 | With the advancement of the unmanned aerial vehicles (UAV) remote sensing technology and the recent relaxation of low-altitude space control in China, UAV remote sensing is expected to find its application in various fields with potentially a huge commercial market. Conventionally UAV images are processed by photogrammetric technologies which are usually of time-consuming, high cost, heavy human intervention albeit high geometric accuracy. For some cases, due to the inherent nature of UAV images, the traditional photogrammetric approaches are unsuitable for its processing. Currently with the development of the material science and control technology, the lightweight UAV platform technology is relatively mature, however how to efficiently process voluminous UAV images has become a bottleneck for its real applications. This thesis is focused on applying hugely successful scene reconstruction technology in computer vision field to scene reconstruction from UAV images. Our main contributions include: (1) Scene reconstruction based on view clustering via camera auxiliary information: One of the most efficient ways to tackle the scalability problem in large scene reconstruction is to break apart the scene into a number of sub-problems, then reconstruct each sub-problem independently, and merge the partial reconstructions at the end. Image clustering without any camera or scene prior information is a difficult problem per se in 3D reconstruction, it is inherently time consuming, and generally no satisfactory results could be achieved. Nowadays, cameras are commonly equipped with crude GPS and compass devices, by which some useful but non-accurate prior information on the image is available. It is expected the image clustering in 3D scene reconstruction could be substantially simplified by taking into account such available but often neglected auxiliary information. We have explored the feasibility and effectiveness of camera auxiliary information based image clustering and cluster-based scene reconstruction. In particular, a view-overlap based clustering approach is proposed, tested and validated by real experiments. (2) Batch reconstruction from UAV images with prior information: After a thorough analysis on the existing problems of the Bundler, a popular increment reconstruction technique in computer vision, for the scene reconstruction of UAV images, a batch reconstruction method is proposed by fully taking into account various pieces of prior infor... |
| 语种 | 中文 |
| 其他标识符 | 200818014628036 |
| 源URL | [http://ir.ia.ac.cn/handle/173211/6525]  |
| 专题 | 毕业生_博士学位论文 |
| 推荐引用方式 GB/T 7714 | 郭复胜. 无人机图像的三维重建方法研究[D]. 中国科学院自动化研究所. 中国科学院大学. 2013. |
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。