摘要
针对钻井现场视频监控人工值守工作量大、异常行为识别不及时、海量视频数据缺乏实时报警等问题,结合钻井作业操作规程建立井控坐岗监测、起下钻、甩钻具等钻井作业场景典型异常行为判别规则,在基于卷积神经网络的目标检测方法基础上,引入SIFT方法和基于强化学习的困难样本筛选策略,提高了算法识别的准确率与鲁棒性。测试结果表明,在像素分辨率1 920×1 080的情况下,该方法整体识别准确率达到85%以上,平均识别速度为135 ms/帧。在此基础上,开发了钻井过程人员异常行为智能识别系统,试点应用结果表明,该系统能够对接入的视频流进行实时检测,及时、准确识别出坐岗人员擅自离岗、起下钻及甩钻具过程人员危险站位等异常行为,有效提升钻井现场安全管理水平。
To solve the problems of heavy workload,untimely abnormal behavior recognition and lack of real-time alarm of video surveillance in drilling field,the recognition rules and detection methods of typical abnormal behavior in drilling operation scenarios were established in accordance with drilling operation procedures,concluding overflow sitting monitoring,tripping and lifting drilling strings. On the basis of convolutional neural network,SIFT method and reinforcement learning based difficult sample selection strategy were introduced into the target detection model to improve the accuracy and robustness. The test results showed that the recognition accuracy of the method was over 85% and the average recognition speed was 135 ms/frame under the condition of 1920 * 1080 pixel resolution. The video intelligent recognition system for abnormal behavior in drilling operations was developed. The application results showed that the system could perform the real-time video stream,accurately identify the abnormal behaviors mentioned above,which could effectively improve the safety management level of the drilling site.
作者
王文正
吴德松
李千登
Wang Wenzheng;Wu Desong;Li Qiandeng(SINOPEC Research Institute of Safety Engineering,Shandong,Qingdao,266071)
出处
《安全、健康和环境》
2020年第2期15-20,共6页
Safety Health & Environment
基金
国家重点研发计划(No.2018YFC0809300)
中国石化科技部项目资助。
关键词
钻井作业
异常行为识别
视频智能识别系统
目标检测方法
钻井安全管理
drilling operation
abnormal behavior intelligent recognition
video intelligent recognition system
target detection method
drilling safety management
