A comparison of tools and techniques for stabilising unmanned aerial system (UAS) imagery for surface flow observations

Abstract. While the availability and affordability of unmanned aerial systems (UASs) has led to rapid development remote sensing applications in hydrology hydrometry, uncertainties related such measurements must be quantified mitigated. The physical instability UAS platform inevitably induces motion acquired videos can have a significant impact on accuracy camera-based measurements, as velocimetry. A common practice data preprocessing is compensation platform-induced by means digital image stabilisation (DIS) methods, which use visual information from captured – form static features first estimate then compensate for motion. Most existing approaches rely either customised tools developed in-house, based different algorithms, or general purpose commercial software. Intercomparison purposes that could serve basis selecting particular tool given conditions not been found literature. In this paper, we attempted summarise describe several freely available DIS applicable total seven six aimed specifically at velocimetry one software were investigated terms their (1) various conditions, (2) robustness, (3) computational complexity, (4) user experience, using three case study with flight ground conditions. an attempt adequately quantify tools, also presented comparison metric root mean squared differences (RMSDs) inter-frame pixel intensities selected features. most apparent between regards method identifying videos, i.e. manual selection automatic. State-of-the-art methods automatic require fewer user-provided parameters are able select significantly higher number potentially (by orders magnitude) when compared identification This allows former achieve accuracy, but feature demonstrated lower complexity better robustness complex field paper does intend identify optimal UAS-based purposes, it aim shed light details implementation, help engineers researchers choose suitable needs specific Additionally, RMSD used order measure velocity estimation uncertainty induced