基于视频智能分析的水电厂人员不安全行为自动化检测系统  

Automatic Detection System of Unsafe Behavior of Hydropower Plant Personnel Based on Video Intelligent Analysis

作　　者：宋刚伟 袁璞 石静涛 聂磊 SONG Gangwei;YUAN Pu;SHI Jingtao;NIE Lei(State Grid Shaanxi Ankang HydroPower Plant,Ankang 725000,China)

机构地区：[1]国网陕西安康水力发电厂,安康725000

出　　处：《自动化与仪表》2024年第5期113-116,142,共5页Automation & Instrumentation

摘　　要：为保证水电厂的生产生活安全,该文提出基于视频智能分析的水电厂人员不安全行为自动化检测系统。摄像机模块采集区域内视频并发送给资源模块;资源模块存储以及传输数据;自动检测模块获取由资源模块传输的视频两帧间的人员行为前景图像,将其输入至深度置信网络中,输出人员行为安全性判断结果,并返回至资源模块;若检测模块检测到不安全行为,资源模块会将不安全信号传递给报警模块发出报警。通过实验可知,该方法获取的视频采集结果细节全面、清晰,提取的人员行为前景图像的交并比高,可实现人员不安全行为准确检测。In order to ensure the safety of production and life in hydropower plants,an automatic detection system for unsafe behaviors of hydropower plant personnel based on video intelligent analysis is proposed.The camera module collect that video in the area and sends it to the resource module.The resource module stores and transmits data.The automatic detection module obtains the foreground image of personnel behavior between two video frames transmitted by the resource module,inputs it into the deep confidence network,outputs the judgment result of personnel behavior safety,and returns it to the resource module.If the detection module detects unsafe behavior,the resource module will transmit the unsafe signal to the alarm module to give an alarm.Experiments show that the details of the video capture results obtained by this method are comprehensive and clear,and the intersection ratio of the extracted foreground images of personnel behavior is high,which can realize accurate detection of unsafe behaviors of personnel.

关 键 词：视频分析 水电厂 不安全行为 自动化检测 高清摄像头 帧差法 

分 类 号：TP274[自动化与计算机技术—检测技术与自动化装置]