Autonomous Flight Planning for Unmanned Aerial Vehicle based Photogrammetry
As the demand for high quality, visually representative 3D virtual environments increases, 3D content creators are seeking alternative methods to traditional 3D modelling of complex natural scenes. Photogrammetry is one such method that allows for the reconstruction of geometry from a series of photographs. When a photograph is captured, a 3D scene is converted into a 2D image, sacrificing depth. Photogrammetry offers a means to recover this missing depth information, allowing for the 3D reconstruction of a scene from a series of photographs. This enables a 3D representation of the scene to be viewed from any position, not just the photographer’s original viewpoint. Due to the increased level of visual fidelity that it is capable of achieving, photogrammetry is becoming increasingly popular as an alternative method for the 3D modelling of complex natural scenes.
This thesis provides an overview of photogrammetry, from capturing a photo set to reconstructing a 3D scene. The requirements of a successful photo set are outlined as well as the most common capture methods: close-range and aerial photogrammetry. Popular photogrammetry software is then discussed including the algorithms, such as stereoscopy and triangulation, that they utilize to reproduce a 3D scene from a photo set. The most common applications for photogrammetry, including 3D meshes, digital elevation models and point clouds, are also examined. This thesis also provides an overview and discussion on the current state of photogrammetry research, with particular focus on Unmanned Aerial Vehicle (UAV) based photogrammetry and autonomous flight planning. From this it is clear that, while UAV based photogrammetry is currently a popular alternative, the potential benefits of autonomous flight planning remain largely unexplored in this area. The development of a framework that explores the benefits that autonomous flight planning provides to UAV based photogrammetry is outlined, and it’s potential as an alternative method to traditional 3D modelling is evaluated.
The framework comprises of a mobile application that allows for a UAV’s flight path to be pre-determined and an onboard script that enables the UAV to automate it’s flight trajectory based on this data. The flight path is calculated using a scanline fill of the area of the intended real-world environment to be converted to a 3D environment, as specified by a user. Using data collected from this framework, a partial qualitative comparative analysis of manual and autonomous UAV based photogrammetry is conducted. Findings from the analysis conclude that UAV based photogrammetry clearly benefits from autonomous flight planning, particularly in regards to usability and efficiency. Further analysis would be necessary to evaluate the complete impact that it has on other areas that were unable to be determined by this study. Several new avenues for future research in this area are also briefly discussed, including parallelization and interactive 3D time lapses, which have the potential to further increase the viability of autonomous flight planning for UAV based photogrammetry.