油气田智能视频报警系统设计与实践  

Optimization Design and Application of Intelligent Video Alarm System in Oil and Gas Filed

作　　者：刘湃 耿玉广 马彪 王宇蒙 田彦林 张巍 赵立庆 LIU Pai;GENG Yuguang;MA Biao;WANG Yumeng;TIAN Yanlin;ZHANG Wei;ZHAO Liqing(Engineering Technology Research Institute of Huabei Oilfield Company, Heibei Renqiu, 062552, China;The First exploitation Factory of HuabeiOilfield Company, Heibei Renqiu, 062552, China)

机构地区：[1]中国石油华北油田分公司工程技术研究院,河北任丘062552 [2]中国石油华北油田分公司第一采油厂,河北任丘062552

出　　处：《数码设计》2018年第4期28-32,共5页Peak Data Science

摘　　要：随着油气田开发的不断深入,远程视频监控系统在油田生产管理发挥了较大作用。但是由于生产环境复杂、监控环节众多、报警辨识度低等因素限制,现有的视频监控系统主要应用于人工甄别管理与事后问题分析。本文针对如何提高油气田视频监控系统的报警准确率的问题进行探讨,基于深度学习的SSD图像处理技术,设计了油气田视频智能报警系统。通过现场实验验证,该系统对视频中车辆及人员等的目标辨识度达到90%以上,报警响应速度在2s以内,系统内嵌误报信息自学习技术,极大提高油气田视频监控系统报警成功率及使用率。and Surveillance is playing a more and more important role in the oilfield management. Because the oil production environment is complicated and the workers can’t always monitor the wells, the alarms distinguishing degree decrease. The oilfield monitoring system which provide the real-time image data, is the key part of the modern safe production monitoring system. This paper study on how improve the alarm accuracy of the oilfield video monitoring system. By using the deep learning SSD image processing technique, the oilfield video monitoring system was designed. By using the video monitoring system in the oil transferring station, the target identification reached more than 90%, the alarm response speed was less than 2s. The video monitoring system which can reduce false alarm rate and the missing alarm rate by using the false alarm information self-learning technology, resolved the oilfield video monitoring problems of the high false alarm rate and low using.

关 键 词：深度学习 SSD 视频监控 视频报警 目标辨识度 报警响应速度 

分 类 号：TP277[自动化与计算机技术—检测技术与自动化装置]