基于改进YOLOv8的输送带表面破损检测  

Surface Damage Detection of Conveyor Belt Based on Improved YOLOv8

作　　者：梅秀庄[1] 王星覃 徐刚[1] 王颖[1] Mei Xiuzhuang;Wang Xingtan;Xu Gang;Wang Ying(College of Mechanical Engineering,Inner Mongolia University of Technology,Hohhot 010051,China)

机构地区：[1]内蒙古工业大学机械工程学院,呼和浩特010051

出　　处：《煤矿机械》2025年第4期189-191,共3页Coal Mine Machinery

基　　金：内蒙古自治区直属高校基本科研业务费项目(JY20230094);内蒙古工业大学科研启动金项目(DC2200000929)。

摘　　要：针对煤矿输送带表面产生的破损,采用改进的YOLOv8网络进行检测。首先在YOLOv8网络Backbone中的C2f之后添加高效通道注意力(ECA)模块;其次将Backbone中的标准卷积层替换为深度可分离卷积(DSC)层;最后将损失函数替换为高效交并比(EIoU)。实验结果表明,改进后的网络可高效提取破损特征,减少了网络的计算量,加快了网络训练的收敛速度,最终准确率提升至94.8%。Aiming at the surface damage of coal mine conveyor belt,the improved YOLOv8 network was used to detect it.Firstly,the efficient channel attentio(ECA)module is added after C2f in the Backbone of YOLOv8 network;secondly,the standard convolution layer in Backbone is replaced by a depthwise separable convolution(DSC)layer;finally,the loss function is replaced by efficient intersection over union(EIoU).The experimental results show that the improved network can effectively extract the damage features,reduces the calculation of the network,accelerates the convergence speed of network training,and finally improves the accuracy to 94.8%.

关 键 词：输送带 表面破损 目标检测 YOLOv8 

分 类 号：TD528.1[矿业工程—矿山机电]