Rip current and channel detection using surfcams and optical flow

Abstract

Rip currents are a common, naturally occurring surf-zone hazard that pose a risk to beach patrons. This study presents a remote-sensing-based algorithm to detect rip currents and rip channels. Optical flow-based computer vision methods are implemented to analyze large data sets and the automatic detection of these features. Surfcam video was collected from dissipative (La Jolla, CA), intermediate (Long Beach, NY), and reflective beaches (Pensacola Beach, FL) to demonstrate the efficacy of the methods. A clustering technique using the dominant wave period was implemented to transition from detected offshore movements to rip currents. The methods presented in this paper were used to detect 20,327 rip currents and 1,100 rip channels. The average accuracy for rip current and rip channel detection was 67.3% and 96.2%, respectively. The remote-sensing-based detection methods can be adapted for use on other video-based equipment and, with additional modifications, can be implemented in an operational capacity.

RipViz: Finding Rip Currents by Learning Pathline Behavior

“…Likewise, Rashid et al [35,36] also used timex images but utilized a modified version of the Tiny-Yolo V3 architecture. Similarly, Ellis and McGill [12] used timex images in conjunction with environmental information such as tides, wave height, and period to cluster offshore movements to rip currents. However, their method produced false positives in situations where non-water objects such as surfers, paddle boarders, etc., are also moving offshore.…”