Beach Monitoring

Since 2011, SOCIB has been conducting systematic and continuous monitoring of beaches using cameras, generating valuable long-term time-series data. This data is accessible to scientists, coastal management authorities, and citizens, and is currently utilized for shoreline tracking. In the iMagine project, we aim to develop a prototype service that processes video images from beach cameras to monitor the formation and dismantling of seagrass beach berms (Posidonia oceanica) and detect rip currents.

Analyzing the existing system, we identified gaps and bottlenecks. Currently, the shoreline position is manually extracted from the SIRENA video-monitoring system at infrequent intervals (~15 days), even though images are captured multiple times per day. No other beach features are extracted, despite the potential to obtain valuable information about biogeophysical and socioeconomic processes, such as identifying Posidonia berms, rip currents, determining beach width, swash zone, and run-up. By utilizing Deep Learning (DL) techniques for image segmentation, we can automatically extract information on important features like sand, water, white scum, Posidonia berms, humans, and vessels. This automation would allow for shoreline extraction at almost all available timestamps and characterization of Posidonia berms, among other essential aspects of beach monitoring and management. Additionally, DL applied to object identification could aid in the identification of rip currents, which is crucial for emergency services, forecasting models, and early-warning systems.

Our development roadmap begins with enhancing metadata and selecting appropriate deep learning and labeling methods for data preparation and labeling. We then transfer the training dataset to the iMagine AI platform to commence model development, training, and validation. Once the model is ready, we integrate it into the prototype service.