基于改进YOLOv8n的织物疵点检测方法研究  

Research on Fabric Defect Detection Method Based on Improved YOLOv8n

作　　者：王可 李欣雨 郑彬朋 李曦 宋森 WANG Ke;LI Xin-yu;ZHENG Bin-peng;LI Xi;SONG Sen(School of Computer Science,Xi'an Polytechnic University,Xi'an 710699,China)

机构地区：[1]西安工程大学计算机科学学院,陕西西安710600

出　　处：《电脑与电信》2024年第10期14-20,50,共8页Computer & Telecommunication

基　　金：西安工程大学2023年大学生创新创业训练计划项目,项目编号:S202310709117。

摘　　要：针对YOLOv8n模型在织物疵点检测任务中存在疵点种类识别错误和微小疵点识别率不高的问题,研究提出了一种基于改进YOLOv8n的织物疵点检测算法。该研究在织物疵点检测模型中引入了动态蛇形卷积和SE注意力机制,将YOLOv8n模型的主干网络中的部分C2f模块与动态蛇形卷积相结合,并在主干网络中加入了SE注意力机制,以提高模型的检测效果。实验结果显示:与原始YOLOv8n相比,改进后的算法精确率提升了7.2%,召回率提升了2.1%,mAP50值提高了3.6%,mAP50-90值提高了1%。这表明,基于改进YOLOv8n的织物疵点检测模型在疵点检测能力方面得到了显著提升。Aiming at the problems of incorrect defect type recognition and low identification rate of small defects in the YOLOv8n model for fabric defect detection tasks,this study proposes a fabric defect detection algorithm based on an improved YOLOv8n.In this research,dynamic serpentine convolution and SE attention mechanism have been introduced into the fabric defect detection model.Part of the C2f modules in the backbone network of the YOLOv8n model have been combined with dynamic serpentine convolution,and the SE attention mechanism has been added to the backbone network to enhance the detection performance of the model.Experimental results show that,compared with the original YOLOv8n,the improved algorithm has improved precision by 7.2%,recall by 2.1%,mAP50 by 3.6%,and mAP50-90 by 1%.This indicates that the fabric defect detection model based on the improved YOLOv8n has achieved significant enhancements in defect detection capabilities.

关 键 词：疵点检测 YOLOv8n 动态蛇形卷积 SE注意力机制 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] TS101.97[自动化与计算机技术—计算机科学与技术]