【无人机】【2017.11】工程检查和地理空间制图用小型无人机系统
本文为美国俄勒冈州立大学（作者：Farid Javadnejad）的博士论文，共168页。
携带消费级非测量相机的小型无人机系统（UAS）越来越多地被用来生成高分辨率的三维地理空间数据。无人机的低成本、易操作性、广泛可用性和低空机动能力，以及技术和方法的快速发展，使得基于无人机的摄影测量应用于许多土木工程领域，如可视化、三维制图、进度监测、施工、结构检查、维护、监控等。利用计算机视觉技术、运动结构（SfM）和多视图立体（MVS）技术，可以从安装在UAS上的廉价消费级摄像机重建3D场景。尽管UAS越来越受欢迎，但基于UAS的制图产品的准确性仍然存在重大的研究问题。此外，需要新的工具和方法来管理和处理UAS生成的大量数据。这项研究探索了几种新的方法来解决在基于UAS的摄影测量领域的一些需求。
首先，提出了一种利用热红外（TIR）图像对UAS图像中的三维点云进行补充的新方法。目前，仅从重叠的TIR图像重建精确3D场景或点云存在若干挑战。例如，消费级TIR相机具有相对较低的分辨率和较窄的视野；此外，此类相机通常生成边缘和纹理模糊的图像。提出并评价了一种利用共采集的TIR和RGB图像生成三维TIR-RGB点云的方法。首先，使用RGB图像生成3D点云；然后，使用轴测技术将TIR数据归于3D点云。使用该方法，生成包含RGB和TIR信息的稠密点云。这种方法对许多热工测绘和检测项目，包括热损失检测、结构无损检测和输电电气部件检测，都是有益的。其次，研究探讨了利用基于UAS的摄影测量技术探测地面管线变化的可行性。由于飞行成本低廉，无人机是检查、监视和检测站点随时间变化的理想选择。在研究方法中，通过重复的UAS飞行来获得三维数字地形模型，以检测和测量场景中管道的运动。将结果与使用实时动态（RTK）全球导航卫星系统（GNSS）接收机和全站仪进行的常规地面测量的位移测量结果进行了比较。再次，引入新的稠密点云质量因子（DPQF）作为SfM-MVS稠密点云精度评价的代理指标。针对不同的数据采集和场地条件，采用模拟实验和经验实验对基于图像的三维重建模型精度进行了评估。研究了DPQF与重建误差的空间相关性，并进行了多次实验解释。研究结果表明，DPQF可以为三维点云提供一个有用的附加信息场。最后，本研究提出一个新的网络地理资讯系统工具BridgeDex，用来管理高解析度桥梁检视影像的时间感知序列，例如从掌上数码相机或无人机上的摄影机所撷取的影像。该工具可用于管理和查询桥梁检测图像、桥梁报告和其他相关元数据。这个基于web的原型为用户提供了一个简单的界面，用于查看、平移和放大多年来大量桥梁检查而收集的桥梁图像。该工具为用户提供了一种直观、有组织的桥梁检测数据评估和管理方法。
Small unmanned aircraft systems (UAS)carrying consumer-grade nonmetric cameras are increasingly utilized to generatehigh-resolution 3D geospatial data. Low cost, ease of operation, widespreadavailability and low altitude maneuvering capabilities of UAS, as well as therapid development of technology and methods, make UAS-based photogrammetryapplicable to many civil engineering applications such as visualization, 3Dmapping, progress monitoring, construction, structural inspection, maintenance,and monitoring. Using computer vision techniques, Structure from Motion (SfM)and Multi-View Stereo (MVS), it is possible to reconstruct 3D scenes frominexpensive, consumer-grade cameras mounted on a UAS. Despite the increasingpopularity of UAS, significant research questions remain regarding the accuracyof UAS-based mapping products. In addition, new tools and methods are requiredto manage and process the vast amount of data generated by UAS. This researchexplores several novel approaches to address some of these needs in the fieldof UAS-based photogrammetry.
First, a new approach is developed forsupplementing the 3D point clouds derived from UAS imagery with thermalinfrared (TIR) imagery. Currently, there are several challenges forreconstructing accurate 3D scenes or point clouds solely from overlapping TIRimagery. For instance, consumer-grade TIR cameras have relatively lowresolution and a narrow field view; moreover, such cameras usually generateimages with blurred edges and textures. An approach is proposed and evaluatedfor generating 3D TIR-RGB point clouds utilizing coacquired TIR and RGB images.First, a 3D point cloud is generated using the RGB images; afterward, the TIRdata is attributed to the 3D point cloud using a boresight technique. Using theproposed approach, dense point clouds are generated that contain both RGB andTIR information. Such an approach can be beneficial for many thermal mappingand inspection projects, including heat loss inspection, non-destructivetesting of structures, and electrical parts inspection for power transmission.Second, the research examines the feasibility of utilizing UAS-basedphotogrammetry for detecting the change of above-ground pipelines. Since it isinexpensive to fly, UAS are ideal for inspecting, monitoring, and detectingchanges in the sites over time. In the researched approach, repetitive UASflights were conducted to derive 3D digital terrain models in order to detectand measure the movement of pipelines in the scene. The results are comparedwith displacement measurements taken from conventional ground surveys usingreal-time kinematic (RTK) global navigation satellite system (GNSS) receiversand total stations. Thirdly, this study introduces new dense point cloudquality factors (DPQF) to use as proxy indicators for assessing the accuracy ofSfM-MVS dense point clouds. Simulated and empirical experiments are used toassess the accuracy of image-based 3D reconstructed models with respect todifferent data collection and site condition scenarios. The spatial correlationbetween the DPQFs and the reconstruction error is investigated and interpretedfor multiple experiments. The results of this study show that the DPQF can be ahelpful additional field of information for 3D point clouds. Last, thisresearch introduces a new web GIS tool, named BridgeDex, for management oftime-aware series of high-resolution bridge inspection images, such as imagescollected from handheld digital cameras or cameras on UAS. This tool can be usedto manage and query bridge inspection images, bridge reports, and otherrelevant metadata. This web-based prototype provides the user a simpleinterface for viewing, panning and zooming in and out of bridge imagerycollected over the years as a result of numerous bridge inspections. The toolprovides the user an intuitive, organized method for evaluating and managingbridge inspection data.
- 引言
- 三维点云生成中自然色与热红外UAS图像融合的摄影测量方法
- 基于无人机系统的地面运动监测摄影测量
- 密集点云质量因子（DPQF）作为图像三维重建精度的代理指标
- BRIDGEDEX：用于管理和查询多年多尺度桥梁检测图像的WEB GIS平台
- 结论
更多精彩文章请关注公众号：