基于改进Yolov5的花色布匹瑕疵检测方法  

Fancy Cloth Defect Detection Method Based on Improved Yolov5

作　　者：时造雄 茅正冲[1] SHI Zaoxiong;MAO Zhengchong(School of Internet of Things,Jiangnan University,Wuxi 214000,China)

机构地区：[1]江南大学物联网工程学院,江苏无锡214000

出　　处：《计算机测量与控制》2023年第4期56-62,共7页Computer Measurement &Control

摘　　要：花色布匹的瑕疵检测是纺织工业中必不可少的环节,实现快速、准确的花色布匹瑕疵检测对于提高生产效率具有重要意义;针对花色布匹瑕疵检测中大部分瑕疵目标较小、种类分布不均、部分瑕疵长宽比较为极端以及瑕疵与背景易混淆的检测难点,提出了一种基于YOLOv5网络改进的算法模型DD-YOLOv5;在骨干网络中采用上下文变换器网络(CoTNet),增强视觉表示能力;在颈部网络中引入卷积注意力模块(CBAM),使网络学会关注重点信息;在检测环节增加了一个高分辨率的检测头,加强对小目标的检测;并且使用α-IoU代替原网络中G-IoU方法;经实验证明,改进后的算法在花色布匹瑕疵数据集平均精度均值上(mAP)达到了较原生算法相比提升了8.1%,检测速度也达到了73.6 Hz。The defect detection of fancy cloth is an indispensable link in the textile industry.It is of great significance to realize the rapid and accurate defect detection of the fancy cloth to improve the production efficiency.In order to solve the defect detection difficulties of small defect targets,uneven distribution of types,extreme length and width of some defects,and easy confusion with background,an improved algorithm model DD-YOLOv5 based on YOLOv5 network was proposed.Contextual transformer networks(CoTNet)are used in the backbone to enhance the visual presentation capabilities;By introducing a convolutional block attention module(CBAM)into the neck network,the key information is concerned in the network.A high resolution detector is added in the detection link to strengthen the detection of small targets.In addition,theα-IoU method is used to replace original G-IoU method.The experimental results show that compared with the original algorithm,the mean average precision(mAP)of the improved algorithm increases by 8.1%,the detection speed also reaches up to 73.6 Hz,the proposed method has a good performance in the fancy cloth defect detection.

关 键 词：瑕疵检测 深度学习 CoTNet 注意力机制 交并比 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]