Key points are not available for this paper at this time.
Shoreline detection is a fundamental component of shoreline monitoring, which plays a vital role in coastal zone management. My Khe beach, located along the central coast of Vietnam, has experienced significant coastal erosion in recent years, underscoring the need for timely and accurate shoreline detection and change analysis. Although various techniques and models have been proposed, many still face limitations in precision and computational efficiency. This study presents a deep learning-based approach that integrates image calibration based on multi-source field and camera data, including surveillance imagery, ground control points, beach profiles, and tidal levels – with the YOLOv8-segmentation model for automated shoreline extraction and monitoring. The YOLOv8 model achieved an average positional error of 0.883 meters, and a root mean square error of 1.264 meters, significantly outperforming the widely used U-Net model. Additionally, the YOLOv8-based system demonstrated a high processing speed of 84 frames per second, highlighting its potential for real-time monitoring and future applications such as UAV-based shoreline detection that require rapid processing of large, high-resolution datasets. Over a year-long monitoring period, the system effectively captured shoreline changes, demonstrating its practical applicability for long-term coastal management.
Van et al. (Mon,) studied this question.