Design And Implementation Of The Oilfield Surveillance System Based On Intelligent Video Analysis Technology

Petroleum industry is playing a vital role in the process of industrial modernization.In the modern petroleum production system,massive remote petroleum wells have difficulties in supervision,petroleum theft,petroleum leakage and etc.In order to solve these problems,petroleum drilling enterprises construct video monitoring system.However,the existing monitoring system relies on the subjective judgment of the monitoring personnel and lacks of the automatic analysis function of monitoring video.In this paper,an intelligent video monitoring system for remote petroleum fields is designed and implemented.In this paper,we firstly describe the problems that video monitoring system currently applied in petroleum exploration enterprises in the application process,such as the inability to realize high-speed intelligent application,low alarm sensitivity,high false alarm rate,low detection rate,and poor system availability.Secondly,both domestic and overseas video monitoring system are analyzed,including the current situation and existing problems of related target detection technology development.Based on the research and demand analysis of existing remote oil field of video monitoring system,a designing scheme of intelligent video monitoring system is introduced.Technologies applied in this intelligent video monitoring system implementation process are presented as well,including HD monitor,network transmission,video codec,video transmission,video storage,video analysis and etc.Finally,combined with the actual situation of field equipment in Changqing oilfield,the hardware system of video monitoring was improved,and the modified hardware system was debugged and tested.Compared with the existing video monitoring system,this system has certain advantages in both hardware and software.In terms of hardware,the video encoder selected by the video monitoring system proposed in this paper is based on Linux platform.The Linux platform is more stable and has multiple data interfaces.It can access multi-modal data such as analog images and video images,which can protect the original equipment investment in the production site to the greatest extent.The video encoder also supports a variety of video resolution protocols and the function of disconnection and transmission,which can access the data processing center efficiently.In terms of software,this system realizes the rapid extraction of foreground area in the monitoring screen through modeling method based on LBFuzzyGaussion,reducing the area to be identified and the recognition time.Then through R-FCN network,detection is implemented on thesuspicious foreground area to obtain the category and location of the foreground target.When detecting the subsequent frames,only the neighborhood,foreground area and video edge of the suspect target in the previous frame are detected,so as to further improve the detection efficiency.Finally,the detected foreground target type and movement trajectory were compared to determine whether there was an intrusion.When the intrusion and other behaviors were detected,the alarm was triggered,and the alarm video fragment was saved for subsequent responsibility.In this paper,168 alarm video of 36 monitoring points in Changqing oilfield is adopted as the dataset for experimental analysis.The detection system proposed in this paper is compared with the common intelligent video monitoring system.Through the comparative test,it shows that the system achieved a lower false alarm rate and conformed to the design expectation while ensuring the adequate monitoring accuracy.Therefore,the detection accuracy and miss rate of the system fully meet the actual use requirements of operation sites,and the system has a strong ductility,with great engineering practical application value.