Rip current detection based on uav images and convolutional neural network

Abstract

Rip currents are powerful, narrow seaward flows of water in the surf zone. Rips can appear in any beach environment because they are dependent on wave breaking. Their formations are given by wave characteristics, coastal morphology, and wind. From the beach-safety aspect, rip currents could be considered as one of the most dangerous threats to bathers, pulling swimmers away with a speed of 1 – 2.5 m/s. Hence, it leads to a lot of unfortunate consequences. However, rip currents are often difficult to detect and recognize by naked-eye observation. This paper proposes a new method for detecting rip currents using a convolutional neural network and images collected by an unmanned aerial vehicle. The proposed system achieves an accuracy of 90.19 %  on the testing set with the support of 15 frames per second. The results are a premise for building a real application in coastal monitoring systems to detect rip currents.