UAV photogrammetry ground control reductions using GNSS

Abstract

Unmanned aerial vehicles (UAVs) are now well-established as platforms for photogrammetric data acquisition. Their advantages, particularly over conventional manned aerial platforms, relate to their low cost, ease of use, rapid deployability and low-level flying for the collection of centimetre-level spatial resolution imagery. Coupled with recent innovations in photogrammetry and computer vision, UAVs equipped with consumer grade digital cameras are now frequently used to generate centimetre-resolution and accuracy mapping products, such as dense point clouds, digital elevation models and orthomosaics. Despite the efficiency of UAV data acquisition, the continued need for ground control implementation for photogrammetric image orientation remains a substantial workflow constraint. In addition to the associated costs, ground control must be implemented strategically, and usually extensively, to ensure photogrammetric products meet the accuracy requirements of large scale mapping, which may or may not be possible given constraints of the intended application. This research uses high precision, UAV-based GNSS (Global Navigation Satellite System) positioning techniques to substantially reduce ground control requirements by directly determining UAV image positions with centimetre-level accuracy and precision. The Precise Point Positioning (PPP) technique is applied and can yield centimetre-level planimetric and decimetre height accuracy photogrammetric mapping without GCPs, whilst the height accuracy can be improved to the centimetre-level using a single GCP. Unlike the standard relative GNSS positioning technique, PPP alleviates all spatial operating constraints associated with the installation and use of a local ground-based GNSS reference station, or the need to operate within the bounds of a permanent GNSS reference station network. Such a workflow simplifies operational logistics, and enables large-scale photogrammetric mapping from UAVs in even the most remote and challenging geographic locations globally. The approach was tested on 11 fixed wing UAV datasets, acquired at two sites in Northumberland, north-east England, which had varying ground control configurations. UAV flight durations, meaning time between launch and landing, were 12-42 minutes. It is shown that the main limitation of UAV-based PPP application is the inherent possibility of GNSS cycle slips and limited observation spans that inhibit the convergence of float ambiguity estimates. Although PPP camera position estimates were biased in such cases, GCPs were still minimised due to the retained precision of the PPP camera position estimates and constraints on the image block.