基于YOLOv8和高分遥感的长江口日本鳗鲡苗捕捞网具分布变化监测研究OA

Anguilla japonica fishing is a key driver shaping the sustainability of wild eel resources,so up-to-date information on fishing effort has become an urgent concern for fisheries managers.Conventional remote-sensing techniques,however,fail to provide effective monitoring,because the gear is extremely small and sea conditions are highly complex.To test whether very-high-resolution satellite imagery can be used to track Anguilla japonica fishing dynamics,we applied a YOLOv8 object-detection model to map and monitor the distribution of fishing nets over two consecutive years(2023 and 2024)in a section of the East China Sea east of Chongming Island at the mouth of the Yangtze River.Training on a single-year sample set proved insufficient for reliable gear detection,whereas a hybrid dataset that combined imagery from both years markedly improved recognition accuracy and the model's generalization under complex sea states,yielding precision(P)and mean average precision(mAP@0.50)values above 0.90.The monitoring results reveal a pronounced year-on-year increase in the number of nets deployed in 2024,indicating large inter-annual variability in fishing intensity and underscoring the need for more effective,dynamic fishing-effort monitoring.The hybrid-dataset,deep-learning strategy developed provides a robust technical reference for implementing dynamic monitoring and management of China's Anguilla japonica fishing.