The computer vision techniques and numerical models for the assessment of wave flow during coastal boulder movements

The impact of extreme wave events, such as hurricanes and tsunamis, has resulted in the displacement of large coastal boulders along various coastlines worldwide. Several coastal boulders have been studied along the Mediterranean coasts to assess the wave flow capable of causing their dislodgment. In particular, recent extreme wave events, primarily associated with the occurrence of Mediterranean hurricanes, have led to the dislodgment of multiple boulders under different pre-setting transport conditions (subaerial/submerged, joint-bonded, cliff-edge). These movements have been recorded by surveillance cameras located along the coasts, providing direct evidence of boulder displacements under well-defined transport conditions. A detailed understanding of the pre-setting transport conditions and types of movements (sliding, overturning, saltation) is crucial for evaluating the theoretical wave flow needed to initiate boulder transport. In this study, a comparison was performed between theoretical wave flow and computed wave flow, utilizing numerical models of incipient motion formulas along with the recorded data obtained through computer vision techniques. Morpho-topographic surveys were conducted at different times on a rocky coast in southeastern Sicily (Italy), which experienced boulder dislodgments during various impacts of Mediterranean hurricanes. Terrestrial Laser Scanning (TLS) and Structure from Motion (SfM) techniques were used to assess the dimensional parameters of the coastal boulders. Subsequently, numerical models based on incipient motion formulas were applied to determine the theoretical wave flow required to initiate boulder movement. To compute the wave flow impacting during a given storm event, we applied computer vision techniques to analyze video recordings from surveillance cameras that captured the moments of boulder movement. The video recordings were automatically georeferenced using Ground Control Points extracted from TLS and SfM data to obtain a planar view of the wave propagation with adjusted perspective wrapping. Optical Flow was then applied to the georeferenced video recordings in order to compute the wave flow during these movements. The comparison between computed and theoretical flow provided useful insights for sensitivity analysis of friction, drag, and lift coefficients, thereby improving the accuracy of the force assessments in the incipient motion formulas.