Computational Vision in Photogrammetry for Georeferencing: Modern Resources Evaluation for UAV Image Processing

Abstract

Environmental Science aims to understand the world; this can be achieved by using geographic information systems (GISs). Also, georeferencing allows for the adjustment and alignment of raster data in combination with other GIS data. Thus, it is possible in terms of these techniques to interpret these types of data and their relationships, patterns, and trends. This study aims to investigate the use of modern engineering procedures, known as computational vision in photogrammetric image processing, as obtained from unarmed aerial vehicles (UAV). This is done via a small camera in the front of an embedded system and combined with proprietary software that uses computational vision resources. Although open-source software was the prioritized choice, the research began with a study on state-of-the-art computational vision algorithms and photogrammetry for drone inspection.  The generation, processing, and verification of a set of photographic images were further procedures accompanying the study of the algorithms and photogrammetry which subsequently resulted in a georeferencing system. In fact, proprietary software employing computational vision resources was used in this study at the Universidade de Trás-os-Montes e Alto Douro in Portugal to compare it with the conventional methodology using modern computational resources to determine the benefits achieved. In conclusion, the positional quality of the generated georeferencing system was verified, and satisfactory results were reported. This underscored the potential of these modern computational resources in contemporary photogrammetry.