基于改进YOLOv7的输送带钢丝绳芯断裂检测研究  

Study on Conveyor Belt Wire Rope Core Breakage Detection Based on Improved YOLOv7

作　　者：金鑫 李敬兆[1] 刘泽朝 JIN Xin;LI Jingzhao;LIU Zechao(School of Electrical and Information Engineering,Anhui University of Science and Technology,Huainan Anhui 232001,China)

机构地区：[1]安徽理工大学电气与信息工程学院,安徽淮南232001

出　　处：《兰州工业学院学报》2025年第2期26-31,共6页Journal of Lanzhou Institute of Technology

基　　金：国家自然科学基金资助项目(52374154);安徽理工大学研究生创新基金资助项目(2022CX1008)。

摘　　要：为了解决X射线在钢丝绳芯输送带检测中存在电离辐射,且断裂故障尺度变化大,导致检测人员安全和现有模型精度较低等问题,提出基于太赫兹技术的钢丝绳芯输送带断裂检测方法。首先使用太赫兹波对钢丝绳芯输送带进行扫描成像,并通过直方图均衡化和伽马矫正对扫描后的图像进行增强处理,然后由改进的YOLOv7模型进行检测。YOLOv7的改进直接影响最终的检测结果,改进的YOLOv7融合了EMA注意力机制,同时引入CoordConv模块,并使用Soft-NMS替换传统的非极大值抑制三个方面。实验结果表明,与常见的Faster R-CNN、SSD、YOLOv5等目标检测算法相比,改进后的YOLOv7模型在基于太赫兹扫描的钢丝绳芯输送带断裂图像上检测精度达到了93.1%,其检测精度更高,综合效果更好。To address the issues of ionizing radiation in the detection of steel wire rope core conveyor belt by X-ray,and the fault scale changes greatly,which leads to the safety of inspection personnel and low accuracy of existing models,a new method of steel wire rope core conveyor belt fracture detection based on terahertz technology is proposed.In this way,terahertz waves are used to scan and image the steel wire rope conveyor belt,and the scanned images are enhanced by histogram equalization and gamma correction,and then detected by the improved YOLOv7 model.The improvements to YOLOv7 directly affect the final detection results.The improved YOLOv7 integrates the EMA attention mechanism,introduces the CoordConv module,and replaces the traditional non-maximum suppression with Soft-NMS in three aspects.The experimental results show that compared with common target detection algorithms such as Faster R-CNN,SSD and YOLOv5,the improved YOLOv7 model has a detection accuracy of 93.1%on the broken image of steel wire rope conveyor belt based on terahertz scanning,which is higher in detection accuracy and better in comprehensive effect.

关 键 词：钢丝绳芯输送带 太赫兹技术 YOLOv7模型 EMA注意力机制 CoordConv模块 Soft-NMS 

分 类 号：TD528[矿业工程—矿山机电]