基于YOLOv8改进算法的织物瑕疵检测方法  

Fabric defect detection method based on improved YOLOv8 algorithm

作　　者：张学林 闵悦[2] 熊金泉 丁文超 ZHANG Xuelin;MIN Yue;XIONG Jinquan;DING Wenchao(School of Big Data,Jiangxi Institute of Fashion Technology,Nanchang,Jiangxi 330201,China;School of Fashion Design,Jiangxi Institute of Fashion Technology,Nanchang,Jiangxi 330201,China;Shanghai Stratosphere Information Technology Co.,Ltd.,Shanghai 200232,China)

机构地区：[1]江西服装学院大数据学院,江西南昌330201 [2]江西服装学院服装设计学院,江西南昌330201 [3]上海同温层科技有限公司,上海200232

出　　处：《毛纺科技》2025年第3期145-150,共6页Wool Textile Journal

基　　金：教育部人文社会科学研究项目(22YJA760060);教育部产学研项目(202102001044)。

摘　　要：为了解决织物生产过程中,瑕疵检测存在的准确率低、检测速度慢的问题,提出一种基于YOLOv8改进算法的织物瑕疵检测方法。首先,借鉴轻量化的StarNet重新设计了主干网络结构,降低模型参数量,提升检测速度;其次,设计了一种基于Sobel算子的边缘信息增强卷积,以获取瑕疵的边缘信息,提升瑕疵特征的提取能力;最后,在回归损失函数中引入对不同尺度的物体不敏感的NWD损失函数,提高对小目标瑕疵识别的检测能力。实验结果表明:改进后的YOLOv8算法平均检测精度较原模型提升1.5%,模型计算量较原模型下降10.59%,证明了算法改进的有效性。In order to solve the problems of low accuracy and slow detection speed in the defect detection stage of fabric production,a fabric defect detection method based on improved YOLOv8 algorithm was proposed.First,the backbone network structure wss redesigned using the lightweight StarNet to reduce the number of model parameters and improve the detection speed.Secondly,an edge information enhancement convolution based on Sobel operator was designed to obtain the edge information of defects and improve the extraction ability of defect features.Finally,the NWD loss function,which is insensitive to objects of different scales,was introduced into the regression loss function to improve the detection ability of small target defect recognition.Experimental results show that the average detection accuracy of the improved YOLOv8 algorithm is 1.5% higher than that of the original model,and the calculation amount of the model is 10.59% lower than that of the original model,which proves the effectiveness of the improved algorithm.

关 键 词：YOLOv8 瑕疵检测 StarNet SOBEL算子 NWD损失函数 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]