摘要
在当今信息化智能化快速发展的时代，航空航天摄影测量与遥感技术作为获取地球表层定位信息和语义信息的重要手段得到了极大发展，在此背景下，近年来可获取测绘地理信息产品的遥感影像（包括国产高分辨测绘卫星影像、无人机航空影像等）也得到广泛应用。数字高程模型（digital elevation model，DEM）作为4D（DEM、DOM、DLG、DRG）产品之一，在测绘、国土和国防建设等领域有着广泛的应用。然而利用传统摄影测量—影像匹配技术获取的点云数据，在提取DEM时仍存在滤波算法自动化程度低，算法面向不同地形时通用性差，在生产DEM后处理中需花费大量人力物力等问题。针对上述问题本文对航空影像匹配点云滤波进行了研究。另外，结合我国现阶段正在实施的具体项目（第三次国土调查、全球测图等），如何对成果DEM进行滤波处理，以高效解决生产DOM中所出现的地物扭曲、拉花等问题，本文也进行了作业模式的探索并实践生产。据此，本文主要研究内容和结论如下： （1）针对航空影像匹配点云，该文提出一种融合多源数据的影像匹配点云滤波算法。首先融合影像和匹配点云高程多源信息；再引入新型分类器对融合影像进行分类，其分类结果作为知识引导用于点云滤波中，即将分类专题图与影像匹配点云叠加以过滤非地面点，实现点云滤波并生成数字高程模型；最后选用航空影像进行密集匹配和滤波试验。实验结果表明，利用该点云滤波算法生成的DEM与参考DEM呈现高度相关性，可大大减少生产DEM人工后处理的工作量。
陶瓷喷嘴（2）选取航空影像匹配点云采用最新布料模拟技术进行匹配点云滤波。结合该算法原理即在点云翻转的基础上，借助一块符合弹性定律的布料，在内力和外力共同作用下，通过阈值对比布料与点云之间的距离达到滤波目的。经试验，CSF算法对不同地形场景的影像匹配点云可以获取较好的滤波效果，对比传统算法在滤波精度和算法效率上均有优势。
（3）国土调查中数字正射影像（DOM）底图生产是开展资源与环境监测、国情普查的基础。针对正射纠正后影像中存在的大量扭曲区域，特别是道路区域，本文提出全新作业模式，即通过滤波处理道路扭曲区域DEM，进行实时局部正射纠正来实现。高分辨测绘卫星影像高分二号和北京二号两组实验数据生产表明，该方法实现了影像道路扭曲区域的快速纠正，且纠正后道路区域边缘地物没有出现明显错位现象。另外，通过一定数量道路区域控制点，对比局部纠正前后精度，也表明该方法很大程度提高了道路扭曲区域的纠正精度，满足相关正射影像生产精度要求。
综上，本文针对航空影像匹配点云，研究了面向生产的DEM滤波方法及其相关问题；结合具体生产需求，通过对测绘卫星的成果DEM进行滤波处理，使其应用于DOM 典型地物纠正中，并给出高效作业模式和实践生产。
关键词：DEM；点云滤波；影像匹配；布料模拟；国产测绘卫星；道路扭曲
论文类型：应用研究
Abstract
输液恒温器In the era of rapid development of information and intelligence, aerospace photogrammetry and remote sensing technology has been greatly developed as an important means to obtain location and semantic information of the earth surface. In this context, remote sensing images (including domestic high-resolution mapping satellite images, UAV images, etc.) that can be obtained for surveying and mapping geographic information products have also been widely used in recently years. The Digital Elevation Model (DEM) is one of 4D (DEM, DOM, DLG, DRG) products and has been widely used in fields such as surveying and mapping, land and national defense construction. However, the point cloud data obtained by the traditional photogrammetry—images matching technology still has a poor levels of automation when extracting DEM. The algorithm has poor generality when dealing with the different terrains, and it takes a lot of manpower and material resources in the DEM post-processing. In the above question, this paper studies the aerial image matching point cloud filtering. In addition, combined with the specific project being implemented in china at this stage (the third national survey, global mapping, etc.), how to apply product of DEM filtering to effectively solve the problems of distortion and smear in the production DOM, this paper also explores the operation mode and puts it into practice. This main research contents and conclusions of this paper are as follow:
(1)Aiming at the aerial image matching point cloud, a new method was proposed for point cloud filtering. Firstly, it utilizes an image classification method that combines multiple source information and introduces a new classifier, and then knowledge guidance of using its classification results is used in point cloud filtering, that is to say, removing non-ground points is carried out by overlapping a thematic map of classification with a point cloud. Finally, point cloud filtering is implemented and obtain the DEM. The Aerial image is used for an experiment that include dense image matching and filtering test. The Aerial image is used for an experiment that include dense image matching and filtering test. The quantitative evaluation results show that the method can reduce the manual post-processing by a long way and improve the DEM accuracy further more.
(2)Aerial images were selected to match the point cloud and this paper employed latest cloth simulation technology for point cloud filtering. Combined with the principle of the algorithm, that is on the basis of point cloud turning upside down, and with a piece of cloth that conforms to the law of Hooke's law and is affected by internal and external forces, filtering was done by comparing threshold distance between cloth and point cloud. Test was carried out and the results showed that the CSF algorithm can obtain better filtering effects for point clouds of different terrain scenes. Compared with traditional algorithms, it has advantages in filtering precision and algorithm efficiency.
(3)It is the basis for resource and environmental monitoring and census of national conditions when orthophoto is produced as the base map in land survey. Aiming at the large number of distorted areas existing in the orthorectified image, especially the road area, this paper proposes a new method. That is to say, Real-time local orthorectification is carried out by editing DEM which is in the road distortion area. It is shows that the method achieves a rapid rectify of the image road distortion area, and there is no obvious misalignment of the road area of the edge after rectify by the high-resolution satellite imagery of the GF-2 and TRIPLESAT-2 experimental data. In addition, through a certain number of control points in the road area, comparing the accuracy of Local orthorectification before and after, it also shows that the method significantly improves the rectify accuracy of the road distortion area, and meeting the requirements of relative Orthophoto production accuracy.
In summary, aimed at the aerial image matching point cloud, this paper studies the DEM filtering method for production and its related problems. Combined with the specific production requirements, the application of mapping satellite DEM in the typical ground feature correction of DOM is given, and an efficient mode of operation was given and puts it into practice.
Key Words：DEM; point cloud filtering; image matching; cloth simulation; road distortion地下水位监测
目录
摘要..................................................................................................................................... I Abstract .................................................................................................................................... III 1 绪论. (1)碳管炉
1.1 研究背景及意义 (1)
1.1.1 背景介绍 (1)
1.1.2 研究意义 (2)
1.2 研究现状 (2)
1.2.1 影像匹配技术研究现状 (2)
1.2.2 LiDAR点云滤波算法研究现状 (3)
1.2.3 影像匹配点云滤波算法研究现状 (4)
1.3 论文组织结构 (5)
2 影像匹配及点云滤波基础 (6)
2.1 影像匹配技术 (6)
2.1.1 匹配基元 (6)
bttt2.1.2 约束条件 (7)
2.1.3 优化过程 (8)
2.1.4 匹配策略 (9)
2.2 两种点云数据 (9)
abp2632.2.1 机载LiDAR点云 (9)
2.2.2 影像密集匹配点云 (10)
2.2.3 LiDAR点云与影像匹配点云对比 (11)
2.3 图像滤波与点云滤波 (12)
2.3.1 空间域滤波 (12)
2.3.2 频率域滤波 (14)
2.3.3 点云滤波 (15)
3 基于多源数据融合的航空影像匹配点云滤波方法 (17)
3.1基于多源数据融合的影像匹配点云滤波 (17)
3.1.1 影像匹配点云及其深度图获取 (18)
3.1.2 多源数据融合和影像分类 (19)
豫公网安备41160202000603
站长QQ:729038198
关于我们
投诉建议