Effective monitoring of harmful algal blooms (HABs) is crucial for the protection of marine ecosystems, the preservation of biodiversity, and the safeguarding of coastal economies. Phaeocystis globosa, as a widely distributed and ecologically harmful bloom-forming species, warrants particular attention. This study, taking the winter 2023-2024 Phaeocystis globosa bloom event in Xiamen Bay, China as an example, proposes an innovative method based on a high-resolution video surveillance system mounted on buoys. By leveraging the video data collected through this system and utilizing the YOLOv8 object detection algorithm, the real-time monitoring of the dynamics of Phaeocystis globosa was achieved. Compared to traditional methods such as satellite remote sensing or manual sampling, buoy-based video surveillance offers several advantages, including continuous data collection, higher spatial resolution, and real-time detection of dynamic environmental changes. The collected images were processed using deep learning techniques, enabling accurate identification and quantification of algal blooms. During the Phaeocystis globosa bloom event in Xiamen Bay in winter 2023, over 80,000 surface algal targets were annotated, and a YOLOv8 detection model was trained. The trained model demonstrated excellent performance, achieving an accuracy of 0.96, precision of 0.97, and recall of 0.99. By integrating real-time video surveillance with deep learning models, this method significantly improves the accuracy and responsiveness of Phaeocystis globosa monitoring, providing an innovative real-time solution for harmful algal bloom detection. This approach holds certain benefits compared to traditional methods and provides valuable support for early warning and ecological protection in response to HABs.

Phaeocystis globosa, YOLOv8, harmful algal blooms, video surveillance, deep Learning