基于双目视觉的桥式起重机吊钩定位系统研究  

Research on bridge crane hook positioning system based on binocular vision

作　　者：高明科 陈薇 丁勇[1] GAO Mingke;CHEN Wei;DING Yong(School of Safety Science and Engineering,Nanjing University of Science and Technology,Nanjing 210094,China;Nanjing Vocational University of Industry Technology,Nanjing 210024,China)

机构地区：[1]南京理工大学安全科学与工程学院,江苏南京210094 [2]南京工业职业技术大学,江苏南京210024

出　　处：《现代电子技术》2025年第6期147-153,共7页Modern Electronics Technique

基　　金：国家重点研发计划资助项目(2022YFC3005502);国家自然科学基金长江水科学研究联合基金项目(U2240221);国家自然科学基金资助项目(51979174)。

摘　　要：在工业生产现场,起重机械被广泛用于物料搬运,在运行过程中其吊钩对人员安全和财产会造成潜在的损害。为防止这些事故的发生,在基于双目视觉的基础上,提出一种YOLOv5小目标检测算法,通过识别物添加和双目视觉结合的方法实现了起重机吊钩的快速、精确定位。同时,在OpenCV和Python环境下分别对YOLOv5的4种经典网络模型进行训练和预测,并进行了现场实时数据采集、误差分析与校正。研究结果表明,YOLOv5s模型的单帧检测时间为0.15 s,mAP值达到了99.3%,完全满足现场实时响应和定位精度的要求。In industrial production sites,crane machinery is widely used for material handling,and the potential damage to personnel safety and property is caused by its hooks during operation.In order to prevent these accidents,a YOLOv5 small target detection algorithm based on binocular vision is proposed.By combining the addition of objects and binocular vision,the fast and accurate positioning of the crane hook is realized.Four classical network models of YOLOv5 were trained and predicted respectively in OpenCV and Python environments,and on-site real-time data acquisition,error analysis and correction were carried out.The research results show that the single frame detection time of YOLOv5s model is 0.15 s,and the mAP value can reach 99.3%,which fully meets the requirements of real-time response and positioning accuracy on site.

关 键 词：桥式起重机 起重机吊钩 双目视觉 YOLOv5 小目标检测 目标定位 

分 类 号：TN407-34[电子电信—微电子学与固体电子学] TP311[自动化与计算机技术—计算机软件与理论]