Surf zone characterization from Unmanned Aerial Vehicle imagery

Abstract

We investigate the issues and methods for estimating nearshore bathymetry based on wave celerity measurements obtained using time series imagery from small unmanned aircraft systems (SUAS). In contrast to time series imagery from fixed cameras or from larger aircraft, SUAS data are usually short, gappy in time, and unsteady in aim in high frequency ways that are not reflected by the filtered navigation metadata. These issues were first investigated using fixed camera proxy data that have been intentionally degraded to mimic these problems. It has been found that records as short as 50 s or less can yield good bathymetry results. Gaps in records associated with inadvertent look-away during unsteady flight would normally prevent use of the required standard Fast Fourier Transform methods. However, we found that a full Fourier Transform could be implemented on the remaining valid record segments and was effective if at least 50% of total record length remained intact. Errors in image geo-navigation were stabilized based on fixed ground fiducials within a required land portion of the image. The elements of a future method that could remove this requirement were then outlined. Two test SUAS data runs were analyzed and compared to survey ground truth data. A 54-s data run at Eglin Air Force Base on the Gulf of Mexico yielded a good bathymetry product that compared well with survey data (standard deviation of 0.51 m in depths ranging from 0 to 4 m). A shorter (30.5 s) record from Silver Strand Beach (near Coronado) on the US west coast provided a good approximation of the surveyed bathymetry but was excessively deep offshore and had larger errors (1.19 m for true depths ranging from 0 to 6 m), consistent with the short record length. Seventy-three percent of the bathymetry estimates lay within 1 m of the truth for most of the nearshore.