倾斜摄影测量在实景三维建模中的关键技术研究

发布时间：2018-06-22 01:05

本文选题：倾斜摄影测量 + 倾斜影像匹配　；参考：《昆明理工大学》2017年硕士论文

【摘要】：伴随着“数字城市”建设的快速推进,作为地理信息框架的基础数据,城市三维模型越来越受到广泛重视。传统的三维建模方法虽然在模型精细度上能够得到很好的控制,但无法满足大范围快速构建城市三维模型的需求。经过多年的探索研究,摄影测量技术凭借其快速高效的数据获取能力,使大范围的城市三维建模成为可能。但是,由于航空摄影获取的是垂直影像,经常会有地物遮挡、拍摄死角等情况发生,这导致建筑物的侧面纹理信息获取不完整,从而影响三维模型的质量。在此基础上衍生了一种新方法,倾斜摄影测量技术,该技术基于搭载在同一飞行平台的多台相机,能够从多视角采集地物影像,同时影像被赋予精确地理信息,这为快速高效构建城市三维模型提供了一种新思路。本文以倾斜摄影测量技术为出发点,研究倾斜摄影测量构建城市实景三维模型的方法,并对其中涉及到的倾斜影像匹配、三维建模流程、模型修复等关键技术进行研究、分析、总结。本文主要研究内容如下:(1)倾斜影像匹配技术研究。重点介绍了基于尺度不变特征的SIFT算法和基于尺度和仿射不变特征的ASIFT算法,并将两种算法应用于倾斜影像匹配实验中,对匹配效果进行对比分析。同时,结合ASIFT算法中相对倾斜度的性质,提出了不同视角倾斜影像间相对倾斜计算方法,通过实验验证了 ASIFT算法匹配效果与多视影像倾斜度之间的关系。(2)实景三维建模方法研究。利用SWDC-5获取的山西某地区倾斜影像作为实验数据,结合ContextCapture实景三维建模软件,进行实验区的实景三维建模,并对建模过程中涉及到的关键技术进行详细阐述,同时对模型成果复杂度和几何精度作评价和分析。(3)模型成果优化方法研究。针对模型成果部分区域存在建筑物扭曲,地面不平整、漂浮物等问题,结合武汉天际航公司的DP-Modeler软件,对模型成果中缺陷明显的目标和区域进行整饰和替换,达到了精细建模的目的。在此基础上总结了一套模型成果的优化方法。
[Abstract]:With the rapid development of the construction of "digital city", as the basic data of geographical information framework, urban 3D model has been paid more and more attention. Although the traditional 3D modeling method can control the precision of the model well, it can not meet the requirement of constructing the urban 3D model quickly on a large scale. After many years of exploration and research, photogrammetry technology, with its rapid and efficient data acquisition ability, makes it possible for a wide range of urban 3D modeling. However, because aerial photography acquires vertical images, there is often ground occlusion and dead angle, which leads to incomplete acquisition of texture information on the side of buildings, thus affecting the quality of 3D models. On this basis, a new method, tilt photogrammetry, is derived, which is based on multiple cameras mounted on the same flight platform, which can collect ground object images from multiple angles of view, while the images are given accurate geographic information. This provides a new idea for fast and efficient construction of three-dimensional city model. Based on the tilt photogrammetry technology, this paper studies the method of constructing the 3D model of urban real scene by tilt photogrammetry, and studies and analyzes the key technologies, such as tilt image matching, 3D modeling process, model restoration and so on. Summary The main contents of this paper are as follows: (1) Research on tilt image matching technology. The sift algorithm based on scale invariant feature and the sift algorithm based on scale and affine invariant feature are introduced. At the same time, considering the properties of relative tilt in the sift algorithm, a method for calculating the relative tilt between different angles of view is proposed. The relationship between the matching effect of sift algorithm and the inclination of multi-view image is verified by experiments. (2) the method of 3D modeling of real scene is studied. Using the tilt image obtained by SWDC-5 as experimental data, and combining with the 3D modeling software of ContextCapture, the 3D modeling of real scene in experimental area is carried out, and the key technologies involved in the process of modeling are described in detail. At the same time, the complexity and geometric accuracy of the model results are evaluated and analyzed. (3) the optimization method of model results is studied. In view of the problems of building distortion, uneven ground, floating objects and so on in some areas of model results, combined with the DP-Modeler software of Wuhan Tianji Airlines Company, the targets and areas with obvious defects in the model results are trimmed and replaced. The goal of fine modeling is achieved. On this basis, a set of optimization methods of model results are summarized.
【学位授予单位】：昆明理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：P23

【参考文献】