摘要
压力传感器只能监测压力类危险源,对施工人员不安全行为等非物理量相关危险源难以发挥作用,无法全面准确识别。为提高抽水蓄能电站施工现场危险源识别准确性,降低事故风险,提出基于深度学习的识别方法。通过布设摄像头与传感器采集视频图像以获取现场数据,并将视频图像输入深度学习算法,检测危险区域、锁定危险源位置,再进一步识别危险源并设置预警级别。实验表明,该方法的危险源正确识别数量显著高于2种传统方法,且AUC值高,ROC曲线接近理想左上角,证明其识别的高准确性与稳定性。
Pressure sensors can only monitor pressure related hazards,and are difficult to effectively identify hazards related to non physical quantities such as unsafe behavior of construction personnel,making it difficult to fully and accurately identify hazards.To improve the accuracy of hazard identification at the construction site of pumped storage power stations and reduce accident risks,a deep learning based identification method is proposed for research.By deploying cameras and sensors to capture video images and obtain on-site data,the video images are input into deep learning algorithms to detect dangerous areas,lock the location of dangerous sources,and further identify dangerous sources and set warning levels.The experiment shows that the number of correctly identified hazards using this method is significantly higher than the two traditional methods,and the AUC value is high.The ROC curve is close to the ideal upper left corner,proving its high accuracy and stability in identification.
作者
张家坤
ZHANG Jiakun(Anhui Ningguo Pumped Storage Co.,Ltd.,Xuancheng 242000,China)
出处
《电工技术》
2025年第18期98-99,102,共3页
Electric Engineering
关键词
深度学习
抽水蓄能电站
施工现场
危险源
智能
识别
deep learning
pumped storage power station
construction site
hazard source
intelligence
identification
