基于双目视觉的吊卡识别及其方位检测方法  

Elevator Recognition and Orientation Detection Method Based on Binocular Vision

作　　者：李进付 Li Jinfu(Drilling Technology Research Institute,Sinopec Shengli Petroleum Engineering Co.,Ltd.)

机构地区：[1]中石化胜利石油工程有限公司钻井工艺研究院

出　　处：《石油机械》2024年第4期11-17,共7页China Petroleum Machinery

基　　金：国家重点研发计划“低温钻井关键装备设计及研制技术”(2022YFC2806404)。

摘　　要：针对钻井自动化中的吊卡方位需要人工参与,存在识别定位缓慢和准确性差等问题,建立了基于双目视觉的吊卡自动识别和方位检测系统。通过搭建双目相机视觉检测系统,对获取的吊卡图像进行标定、立体校正和立体匹配,进而获取吊卡图像的深度图。在原YOLOv5s目标检测算法的主干网络中引入卷积注意力机制模块,将要识别的目标即吊卡图像进行增强,进一步提高吊卡的识别准确率,结合深度图计算出吊卡中心位置相对机械手的距离和偏转角度,从而实现自动送管。通过在钻井平台的试验,验证了吊卡识别和检测方法的有效性。所得结论可为钻井平台自动化程度的进一步提高提供技术借鉴。In order to solve the problems such as manual involvement,slow recognition and poor accuracy in elevator positioning in drilling automation,a binocular vision based elevator automatic recognition and orientation detection system was built.By means of building a binocular camera visual detection system,the obtained elevator images were calibrated,stereo corrected,and stereo matched to obtain depth maps of them.A convolutional attention mechanism module was introduced into the backbone network of the original YOLOv5s target detection algorithm to strengthen the target to be recognized,i.e.,the elevator image,further improve the recognition accuracy of elevator,and figure out the distance and deflection angle of the elevator center position relative to the manipulator combined with depth maps,thus achieving automatic pipe feeding.The effectiveness of elevator recognition and detection method was verified through tests on drilling platforms.The conclusions provide technical reference for further improving the automation level of drilling platforms.

关 键 词：吊卡识别 方位检测 双目相机 YOLOv5s算法 卷积注意力机制模块 试验验证 

分 类 号：TE928[石油与天然气工程—石油机械设备]