Off-campus UNL users: To download campus access dissertations, please use the following link to log into our proxy server with your NU ID and password. When you are done browsing please remember to return to this page and log out.
Non-UNL users: Please talk to your librarian about requesting this dissertation through interlibrary loan.
Flight Planning for Autonomous Inspection and 3D Reconstruction of Buildings and Infrastructure Using Unmanned Aerial Vehicle
The sheer number of aging infrastructures in the US put great threats to the safety and health of our society. Per the 2019 bridge report (Association, 2019) by the American Road and Transportation Builders Association, 38% of the bridges in the nation have been identified as requiring repairs. More frequent and thorough inspection services are necessary to address the dire situation. The recent advances in unmanned aerial vehicle (UAVs), also known as drone, have made it an economical and effective tool for routine structural inspection compared to the traditional ground inspection tools. Currently, most aerial inspections are performed in a manual fashion with the UAVs fly within the visual line-of-sight, and the inflight data acquisitions are fully handled by the human operators. Operators’ individual skill level and the battery capacity limit often bring uncertainties to the inspection performance in terms of the inspection quality and the coverage completeness. Several view and path planning methods have been actively studied for the autonomous aerial inspection and reconstruction. However, these planning methods were either only focused on designing the shortest path, or could only found sub-optimal results, or were limited to specific inspection scenarios. In this thesis we proposed flight planning methods for autonomous aerial inspection and photogrammetric 3D reconstruction of buildings and infrastructures. Novel viewpoint configurations and optimization strategies were developed in the proposed planning methods to achieve improved inspection and reconstruction results. The first part (Chapter 3) of the thesis is focused on computing the aerial inspection trajectory with a baseline observation quality while follows the constraints of UAVs motion dynamics. This leads to a coverage path planning problem that is solved with a novel view and path co-optimization framework. The framework led to better results with significant improvement of the inspection quality with minor sacrifice of inspection efficiency when compared to state-of-arts. The second part (Chapter 4) provides a two-phase multi-view stereo (MVS) planning approach for aerial photogrammetry of watertight geometries where only exterior surfaces of the objects need to be surveyed, such as buildings. This approach computes the optimal camera views based on a local topological graph which utilizes the property of structure-from-motion (SfM) for images triangulation. Three synthetic scenes, including a residential building, a commercially building and a historical site, were selected to evaluate the proposed approach. The results showed that the method takes less than ¼ of the computation duration while demonstrating comparable or better reconstruction results when compared to the existing approaches. In the third part (Chapter 5), an automated flight planning method for detailed inspection and reconstruction of infrastructures with non-watertight geometries, where both exterior and interior surfaces need to be surveyed such as truss bridges, transmission towers and pipe systems, is proposed. Unlike the existing approaches that sets the network size as an input constraint, the proposed method incorporates a novel camera pose selection mechanism that relies on the domain knowledge of the structure-from-motion to properly determines the size of the optimal camera network. A synthesis truss bridge is selected as the case study to demonstrate the practicability of the proposed method. In conclusion, the proposed approaches addressed some key limitations in the existing autonomous flight planning methods for aerial inspection and photogrammetric reconstruction of buildings and infrastructures. For structure with watertight geometries, the proposed MVS planning method computes the optimal reconstruction results while significantly reduces the processing duration compared to the state-of-arts. For structure with non-watertight geometries, the proposed method is able to sufficiently reconstruct structure details with controllable image size, which are often considered as the fixed constraint in the existing approaches that either results in incomplete surface reconstruction or over-redundant images acquisition.
Shang, Zhexiong, "Flight Planning for Autonomous Inspection and 3D Reconstruction of Buildings and Infrastructure Using Unmanned Aerial Vehicle" (2019). ETD collection for University of Nebraska - Lincoln. AAI27668207.