摘要
为了防止电网基建施工现场发生意外事件,保障施工安全运行,使用数据解析算法识别电网基建施工现场风险事件。利用3D卷积神经网络提取电网基建施工现场视频监控行为语义,通过非监督学习的递归自编码器提取视频的潜在语义特征。采用八叉树形式构建电网基建施工现场静态场景地图进行语义标注,生成电网基建施工现场语义事件描述。将图像语义事件映射到事理图谱中,通过转移概率推理事件发生的风险,实现电网基建施工现场风险事件识别。结果表明:所提方法识别出的图谱可以清晰地展示施工现场风险情况,且识别时间不超过50 ms,准确率在90%以上,且识别清晰程度为98.5%。由此可知,基于数据解析算法可有效识别电网基建施工现场的多种施工风险行为,提高现场施工安全。
In order to prevent risk events from occurring on the construction site of power grid infrastructure and ensure safe op-eration of construction,data analysis algorithms are used to identify risk events on the construction site of power grid infra-structure.This paper utilizes 3D convolutional neural networks to extract the semantic features of video monitoring behavior in power grid infrastructure construction sites,and uses unsupervised learning recursive autoencoders to extract potential semantic features of videos.This paper constructs a static scene map of the power grid infrastructure construction site in the form of an octree for semantic annotation,and generates a semantic event description of the power grid infrastructure construction site.Image semantic events are mapped into the event graph,and risk events are identified on the construction site of power grid in-frastructure by transferring probability inference of the occurrence of events.Testing results show that the graph identified by the proposed method can clearly display the risk situation of the construction site,with a recognition time of no more than 50 ms,an accuracy rate of over 90%,and a recognition clarity of 98.5%.From the above results,it can be seen that data analysis algorithms can effectively identify various construction risk behaviors on power grid infrastructure construction sites,and im-prove on-site construction safety.
作者
刘兵兵
闫丽飞
谷良
宫鑫
何壮壮
LIU Bingbing;YAN Lifei;GU Liang;GONG Xin;HE Zhuangzhuang(Information and Communication Branch,State Grid Shanxi Electric Power Company,Taiyuan 030000,China;Fujian Yirong Information Technology Co.,Ltd.,Fuzhou 350003,China;Taiyuan Power Supply Company of State Grid Shanxi Electric Power Company,Taiyuan 03700l,China)
出处
《微型电脑应用》
2025年第5期101-106,111,共7页
Microcomputer Applications
基金
国网山西电力2023年非结构化平台运营支撑项目资助(XSHT-20230918093900668)。
关键词
数据处理
视频数据解析
电网基建
施工现场
风险识别
场景地图
事理图谱
data processing
video data analysis
power grid infrastructure
construction site
risk identification
scene map
event graph
