机载LiDAR点云与航空影像自动配准的精度分析

发布时间：2018-02-20 07:10

本文关键词： 航空影像机载LiDAR点云自动配准精度分析　出处：《兰州交通大学》2013年硕士论文　论文类型：学位论文

【摘要】：目前为止，航空摄影测量的解析理论已经基本完备，是测绘生产单位制作地图的重要手段之一，经过航空摄影测量一系列处理，可以得到影像的方位元素，还原影像拍摄的位置和姿态，进而生产数字高程模型（Digital Elevation Model，DEM），数字线划地图(Digital Line Graphic，DLG)，数字正射影像图（Digital Orthophoto Map，DOM）和三维城市模型等。机载LiDAR（Light Detection And Ranging，机载激光扫描）系统可以快速获取大面积地表离散点三维坐标数据和其他信息。LiDAR点云密度越来越大，精度也越来越高，其应用推广很有可能引起测量技术的又一次革命。然而，虽然激光扫描能够直接获得目标的三维空间信息，地物分类信息也比较容易获得，但是，它却难以表达目标表面的纹理信息。如果仅仅利用数字影像恢复目标的几何信息需要经过复杂的数据处理流程，涉及到影像特征提取、影像匹配、影像交会、构网及纹理映射等关键问题。激光扫描数据和航空影像数据，在信息的表达、恢复目标的三维信息处理过程中，各有优势和不足，可以互为补充。一方面，影像数据可以弥补LiDAR点云数据纹理表现上的缺陷；另一方面，利用LiDAR点云数据也可以提高影像恢复目标三维几何信息的效率。本文以传统的空中三角测量为基础，按照POS（Positionand OrientationSystem，定位定向系统）辅助空三的作业流程进行研究。首先，研究针对带POS的航空影像，在LiDAR点云数据的辅助下进行空三连接点自动提取；其次，为了获取良好的像控点，参照像控点的布设方案，选取测区局部区域进行基于物方约束的多视影像匹配，得到匹配点云；然后，研究利用ICP算法将匹配点云与机载LiDAR点云配准，并筛选配准点作为像控点，进行平差，经过迭代解算影像在LiDAR坐标框架下的外方位元素，实现航空影像与机载LiDAR数据的自动配准；最后，对配准的精度进行评价，方法是利用专业LiDAR观测软件提取检查点三维坐标，同时在立体观测模式下精确地量测影像上同名点的三维坐标，应用点位中误差的思想分析自动配准精度。在对机载LiDAR点云数据和航空影像的自动配准结果进行精度评价的过程中，涉及的重点与创新点包括以下几方面：（1）利用POS和LiDAR数据在物方约束下进行航空影像空三转点；（2）基于多目视觉在POS和LiDAR数据物方约束下的高精度影像匹配，获取局部区域密集匹配点云，将航空影像与LiDAR点云的配准问题，转化为影像匹配点云与LiDAR点云的配准问题；（3）应用ICP（Iterative Closest Point，迭代最近点）算法进行影像匹配点云与LiDAR点云的配准；（4）航空影像控制片选取，配准精度评定；（5）创新性地提出利用良好配准点作为航空影像的像控点，代替像控点的人工选取，并且提高了像控点的量测精度，以及影像方位元素的解算精度。本文研究的内容，能够较好地解决航空影像与机载LiDAR点云数据的自动配准问题，并且通过评判自动配准精度保证了自动配准的质量，是航空影像数据与机载LiDAR数据集成的关键技术之一，有利于推动航空摄影测量和机载LiDAR技术的更广泛应用。将本文研究的内容集成到机载LiDAR数据处理软件（如：LiDARStation.Pro）和航空影像数据处理软件（如：PixelGrid）等中去，，在应用到测绘生产单位后，能够提高相关数据的处理效率，同时取得一定的经济效益。随着经济技术的发展，社会各界对测绘提出了更高的要求，最大限度地发挥航空摄影和机载LiDAR的技术优势，成为一个测绘相关部门的新需求。本文的研究内容是航空摄影测量数据与机载LiDAR数据集成的前提，为进一步获取更高质量的数字产品，特别是真正射影像和三维城市建模，提供理论基础和技术参考。
[Abstract]:So far, analytical theory of aerial photogrammetry has been basically completed, is one of the important means of production of Surveying and mapping production unit map, after a series of aerial photogrammetry processing, image orientation elements can get the position and attitude of restoring the image shooting, and the production of digital elevation model (Digital Elevation Model, DEM), digital line map (Digital Line Graphic, DLG), digital orthophoto map (Digital Orthophoto Map, DOM) and the 3D city model. LiDAR (Light Detection And Ranging airborne, airborne laser scanning) system can quickly obtain cloud density and large area surface discrete three-dimensional coordinate data and other information of.LiDAR is larger, and precision the higher the application is likely to cause a revolution in measurement technology. However, although the three-dimensional spatial information of laser scanning can directly obtain the goal, feature points Such information is also relatively easy to obtain, but it is difficult to express the texture information of the target surface. If only the recovery target using digital image geometric information requires complex data processing, related to image feature extraction, image matching, image intersection, key problems of triangulation and texture mapping. The laser scanning data and aerial images the data in the information expression, 3D information processing recovery goals, each has advantages and disadvantages, can complement each other. On the one hand, the image data can make up for the defects of LiDAR point cloud data on the performance of the texture; on the other hand, the utilization efficiency of LiDAR point cloud data can also improve the image restoration target 3D geometric information.
This paper is based on the traditional aerial triangulation, according to the POS (Positionand OrientationSystem, positioning and orientation system) of auxiliary operation process three. First, research on aerial images with POS, empty three connection points in the LiDAR point cloud data automatically extracted; secondly, in order to obtain good control reference point, the layout scheme of image control points, selected test area local region object constraint in multi view image matching based on match point cloud; then, ICP algorithm using the matching point cloud and airborne LiDAR point cloud registration, screening and registration points as control points as, after adjustment, the exterior orientation iterative image elements in LiDAR coordinate frame, realize the automatic registration of aerial image and airborne LiDAR data; finally, to evaluate the accuracy of registration. The method is the use of professional LiDAR software to extract the observation check point three At the same time in the dimensional coordinates, three-dimensional observation mode accurately measured image points the 3D coordinates analysis of automatic registration accuracy by using the point position error of thought.
In the process of accuracy evaluation of airborne LiDAR point cloud data and automatic registration results of aerial images, the key points and innovation points include the following aspects:
(1) using POS and LiDAR data to carry out aerial image empty three turning points under the constraint of the object square;
(2) based on the high-precision image matching under the constraint of POS and LiDAR data objects, we get the local area dense matching point cloud based on multi view vision, and transform the registration problem between the aerial image and LiDAR point cloud into the registration problem of image matching point cloud and LiDAR point cloud.
(3) ICP (Iterative Closest Point, iterative nearest point) algorithm is applied to registration of image matching point cloud and LiDAR point cloud.
(4) the selection of aerial image control film and the evaluation of registration accuracy;
(5) innovatively put forward the use of good registration points as the image control points of aerial images instead of manual selection of image points, and improve the accuracy of image control points and the accuracy of image orientation elements.
The content of this paper, can effectively solve the problem of automatic registration of aerial image and airborne LiDAR point cloud data, automatic registration of the quality assurance and evaluation through the automatic registration accuracy is one of the key technologies of the integration of aerial image data and airborne LiDAR data, is beneficial to promote more widely application of aerial photogrammetry and airborne LiDAR technology. The content of this paper will be integrated into the software of airborne LiDAR data processing (such as LiDARStation.Pro) and aerial image data processing software (such as PixelGrid) and go to mapping units after application can improve the processing efficiency of the relevant data, and obtain certain economic benefits.
With the development of economy and technology, society has put forward higher requirements for surveying and mapping, to maximize the advantages of aerial photography and airborne LiDAR, become a new demand for the relevant departments of Surveying and mapping. The research content of this paper is the premise of aerial photogrammetry data and airborne LiDAR data integration, to further obtain higher digital products the quality, especially the real projective image and 3D city modeling, provide theoretical basis and technical reference.

【学位授予单位】：兰州交通大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：P231;P28

【参考文献】