为了更好地了解海州湾海洋公园秦山岛近岸水下生物和生态系统，针对现有水下监测平台监测效率低的问题，本文设计了一种基于YOLOv5(You Only Look Once version 5)的近岸水下目标监测平台。该监测平台基于B/S(Browser/Server)架构，开发了视频信号的接收及存储模块、天气查询地图模块、目标监测模块和数据展示模块。通过水下定点监测，自建秦山岛近岸海洋生物数据集，训练目标检测模型，在目标监测模块中嵌入训练好的YOLOv5目标检测算法和DeepSORT(Deep Simple Online and Realtime Tracking)追踪算法，使监测平台更加智能、准确、高效。结果表明:该万维网(World Wide Web，WEB)平台运行稳定，具有较强的实用性，对于掌握海岸近岸生态情况和保护生态环境具有积极作用。
Design of Offshore Underwater Target Monitoring Platform
In order to better understand the coastal underwater organisms and ecosystems of Qinshan Island in Haizhou Bay Marine Park,a coastal underwater target monitoring platform based on YOLOv5(You Only Look Once version 5)was designed in this paper,aiming at the low monitoring efficiency of existing underwater monitoring platforms.Based on Browser/Server(B/S)architecture,the monitoring platform has developed a video signal receiving and storage module,weather query map module,target monitoring module and data display module.Through underwater fixed point monitoring,self-built offshore Marine biological data set of Qinshan Island,trained the target detection model,and embedded the trained YOLOv5 target detection algorithm and DeepSORT(Deep Simple Online and Realtime Tracking)algorithm in the target monitoring module.Make the monitoring platform more intelligent,accurate and efficient.The results show that the World Wide Web(WEB)platform runs stably and has strong practicability,which plays a positive role in grasping the nearshore ecological situation and protecting the ecological environment.
WEB platformYOLOv5target monitoringReal Time Streaming Protocol
NETL
NSTL
万方数据
