Wave overtopping detection and intertidal zone delineation using semantic segmentation in coastal scenes

Abstract

This paper proposes a camera-based method to detect wave overtopping events and delineate the intertidal zone from coastal imagery. Based on semantic segmentation, each pixel is classified into predefined categories that convey information about nearshore conditions. To better capture shoreline transitions, the taxonomy is enriched with two additional classes, foam and wet sand. The resulting masks are processed to identify overtopping and delineate intertidal zones: overtopping is detected when sea/foam regions reach predefined critical areas, while intertidal extraction relies on the temporal analysis of long image sequences to capture tidal variability. The approach is validated on a dedicated overtopping dataset and on coastal sequences with manually delineated intertidal zones. Results show improved reliability for overtopping detection compared with a generic baseline and good agreement between the extracted intertidal zones and human supervision across different coastal settings. Overall, the proposed method provides practical support for environmental monitoring and early-warning workflows in coastal risk management.