双目视觉下钻杆接口定位的实现  

Implementation of Drill Pipe Joint Positioning Under Binocular Vision

作　　者：张奇志[1,2] 唐凡懿 Zhang Qizhi;Tang Fanyi(School of Electronic Engineering,Xi'an Shiyou University;Shaanxi Provincial Key Lab of Oil and Gas Well Measurement and Control Technology)

机构地区：[1]西安石油大学电子工程学院 [2]陕西省油气井测控技术重点实验室

出　　处：《石油机械》2024年第10期12-19,73,共9页China Petroleum Machinery

基　　金：陕西省科技攻关重点项目“油气钻机远程交互优化控制虚拟仿真平发”(2020GY-046);西安石油大学研究生创新与实践能力培养计划资助项目“双目视觉下钻杆接口定位的实现”(YCS23214241)。

摘　　要：智能机器人技术的快速发展推动了石油钻机装备自动化和智能化的提升。针对目前石油钻机上卸扣作业中,传统的钻杆接口定位存在困难,无法实现完全自动化的问题,提出了一种基于改进YOLOv5x算法的智能钻杆接口检测方法,并在双目视觉下建立了钻杆接口定位模型。模型基于卷积神经网络提取钻杆接口的图像特征,从而实现自动识别;融合CBAM(Convolution Block Attention Module)注意力机制以提高模型的特征提取与表达能力,引入更大尺度特征图作为小目标检测层,减少了钻杆接口图像中小目标的误检和漏检情况;结合SGBM(Semi-Global Block Matching)算法,实现双目视觉下钻杆接口的精准定位;通过网络训练检验改进算法的性能,并在模拟试验环境下开展了钻杆接口的定位试验。试验结果表明:改进YOLOv5x算法平均精度达到98.6%,比原始YOLOv5x算法提高了3.0%;上下接口平均定位误差分别为7.64和6.56 mm,符合工程误差要求,具有一定的工程应用价值。所得结论可为钻杆接口的准确检测和精准定位提供技术参考。The rapid development of intelligent robot technology has promoted the automation and intelligence of drilling equipment.In make-up and break-out of drill pipes currently,traditional tool joint positioning technique is difficult to apply and not fully automatic.To address this problem,a smart tool joint detection method based on improved YOLOv5x was proposed.Then,a tool joint positioning model was built under binocular vision.The model extracts image features of tool joints based on convolutional neural network(CNN)to implement automatic recognition.It integrates the convolution block attention module(CBAM)to improve the feature extraction and expression ability.It also introduces a larger scale feature map as a small object detection layer to reduce the false detection and omissive detection of small objects in tool joint images.Combined with the semi-global block matching(SGBM)algorithm,it implements precise positioning of tool joints under binocular vision.The performance of the improved algorithm was verified through network training,and a positioning test of the tool joint was conducted in a simulated experimental environment.The test results show that the improved YOLOv5x algorithm has an average accuracy of 98.6%,which is 3.0%higher than the original YOLOv5x algorithm.The average positioning errors of the upper and lower joints are 7.64 mm and 6.56 mm respectively,which meet the engineering error requirements and have certain engineering application value.The conclusions provide technical reference for accurate detection and precise positioning of tool joints.

关 键 词：钻井钻杆接口 注意力机制 立体匹配算法 小目标检测 双目视觉定位 

分 类 号：TE92[石油与天然气工程—石油机械设备]