2020 Winner: Flow-based Detection of Rip Currents

Project Information
Flow-based Detection of Rip Currents
AVIS Research Group / CMPS161
Short Summary Video: https://youtu.be/ZiDTnAjh0Ww

Rip currents are like rivers of fast-moving water that can quickly carry the unwary out to sea. They are not easy to recognize especially to the untrained observer. This paper presents a flow-based method of detecting rip currents from short video clips. First, time-varying velocity fields are obtained from the optical flow (a relative change between two consecutive frames in a scene) of a video clip. We then analyze the velocity fields using different flow visualization techniques including pseudo-coloring, arrow glyphs, streaklines, timelines, pathlines, and flow difference. We find that direct application of these techniques does not yield satisfactory results without first enhancing the rip current signal from other detected ocean surface motion and applying additional filtering to highlight the rip current. Together with pre-processing and modifications of these algorithms, we find that a subset of these methods can provide visual clues of the rip current location. We evaluated these methods against previously annotated results by rip current experts. This resulted with some expert annotations being relabeled based on evidence from these techniques.
  • Issei Mori (Nine)