水利信息化Issue(2)：78-85,8.DOI:10.19364/j.1674-9405.2026.02.012
基于深度学习的水利领域智能视频识别系统设计与实现
Design and implementation of intelligent video recognition system in water conservancy sector based on deep learning
摘要
Abstract
Traditional water conservancy video surveillance systems were limited to basic image acquisition,leading to a risk of serious accidents due to inadequate oversight.To address these issues,an intelligent video recognition system was designed and developed for on-site scenarios in water conservancy projects,including reservoirs,rivers,pumping stations,and gate stations.By utilizing an improved YOLOv5 object detection model integrated with Transformer and self-attention mechanisms,the system established a series of detection models covering water gauge reading,floating objects,road waterlogging,and personnel intrusion,enabling real-time analysis of field video streams.Developed in Java using a microservices architecture,the system comprised two core subsystems:AI-based video analysis and model training.It not only performed real-time AI recognition and automatically triggers alert notifications but also facilitated self-iterative model optimization through continuous sample enrichment,thereby improving recognition accuracy.The research results promote the intelligent upgrading of surveillance systems in the water conservancy industry and provide robust technical support for modernizing water conservancy management.关键词
视频识别/系统设计/深度学习/YOLOv5/Transformer模型/微服务架构/模型训练Key words
video recognition/system design/deep learning/YOLOv5/Transformer model/microservices architecture/model training分类
信息技术与安全科学引用本文复制引用
舒依娜,孙宝贵,冯赵云,夏修萍,张伟超..基于深度学习的水利领域智能视频识别系统设计与实现[J].水利信息化,2026,(2):78-85,8.基金项目
国电南瑞科技项目(5246C5250002) （5246C5250002）
