Please wait a minute...
吉林化工学院学报, 2023, 40(11): 38-44     https://doi.org/10.16039/j.cnki.cn22-1249.2023.11.007
  本期目录 | 过刊浏览 | 高级检索 |
CGB-YOLO:用于检测钢铁表面缺陷的YOLO算法
任洛莹1 ,刘兴德2 ,谢延楠1 ,胡文松1 ,余鹏泽1,孔志成1
1.吉林化工学院 信息与控制工程学院,吉林 吉林 132022; 2.吉林化工学院 机电工程学院,吉林 吉林  132022
CGB-YOLO: A Modified YOLO for Ddetection of Steel Surface Defects
REN Luoying1 ,LIU Xingde2 ,XIE Yannan1 ,HU Wensong1 ,YU Pengze1,KONG Zhicheng1
School of Information and Control Engineering, Jilin Institute of Chemical Technology, Jilin City 132022,China; School of Mechanical and Electrical Engineering,Jilin Institute of Chemical Technology,  Jilin City 132022,China
下载:  PDF (4755KB) 
输出:  BibTeX | EndNote (RIS)      
摘要 

针对YOLOv5对金属表面缺陷小目标过多、检测结果易受背景干扰等问题,提出了一种改进的金属表面缺陷检测算法。通过在主干网络引入坐标注意力机制,提高模型对缺陷的关注度;将主干网络中的一些CBS和C3模块替换为GhostNetV2结构构建轻量级的网络,优化模型的性能和效率;在Neck层采用双向特征融合网络(BiFPN)来增强颈部产生丰富的表征,加深整个网络并重用低层次的特征。最后,广泛的实验结果表明,CGB-YOLO在NEU-DET上的精度达到75.0% mAP,比改进前提高了3.8%。该模型在金属表面缺陷检测中具有较好的综合性能。

服务
把本文推荐给朋友
加入引用管理器
E-mail Alert
RSS
作者相关文章
任洛莹
刘兴德
谢延楠
胡文松
余鹏泽
孔志成
关键词:  表面缺陷检测  YOLOv5  深度学习  特征融合     
Abstract: 

In order to solve the problems of YOLOv5 on the problems of too many small targets on metal surface defects and the detection results are easy to be interfered by background, an improved metal surface defect detection algorithm was proposed. By introducing the coordinate attention mechanism in the backbone network, the model pays attention to defects, and some CBS and C3 modules in the backbone network are replaced with GhostNetV2 structure to build a lightweight network to optimize the performance and efficiency of the model. A bidirectional feature fusion network (BiFPN) was used to enhance the neck layer to generate rich representations, deepen the whole network and reuse low-level features. Finally, extensive experimental results show that the accuracy of CGB-YOLO on NEU-DET reaches 75.0% mAP, which is 3.8% higher than that before the improvement. The model has good comprehensive performance in metal surface defect detection.

Key words:  surface defect detection    YOLOv5    deep learning    feature fusion
               出版日期:  2023-11-25      发布日期:  2023-11-25      整期出版日期:  2023-11-25
TP391.4  
引用本文:    
任洛莹 , 刘兴德 , 谢延楠 , 胡文松 , 余鹏泽, 孔志成. CGB-YOLO:用于检测钢铁表面缺陷的YOLO算法 [J]. 吉林化工学院学报, 2023, 40(11): 38-44.
REN Luoying , LIU Xingde , XIE Yannan , HU Wensong , YU Pengze, KONG Zhicheng. CGB-YOLO: A Modified YOLO for Ddetection of Steel Surface Defects . Journal of Jilin Institute of Chemical Technology, 2023, 40(11): 38-44.
链接本文:  
https://xuebao.jlict.edu.cn/CN/10.16039/j.cnki.cn22-1249.2023.11.007  或          https://xuebao.jlict.edu.cn/CN/Y2023/V40/I11/38
[1] 余鹏泽 , 刘兴德 , 谢延楠 , 任洛莹 , 孔志成 , 胡文松. MSP-YOLO:用于识别水果的改进算法[J]. 吉林化工学院学报, 2024, 41(7): 18-25.
[2] 王影, 孙可欣, 刘振刚, 高康盛, 刘麒. 基于改进YOLOv5的仿生机器鱼目标检测算法研究[J]. 吉林化工学院学报, 2024, 41(5): 22-29.
[3] 张天驰, 于紫南. 基于自定义深度学习网络双馈风机接入弱电网故障识别[J]. 吉林化工学院学报, 2024, 41(5): 54-59.
[4] 王敏. 自适应学习中基于CNN和IIDLA的图像识别方法研究[J]. 吉林化工学院学报, 2024, 41(3): 56-61.
[5] 王辅民, 周红娟, 冯国亮, 邢雪. 基于特征融合的空压机故障诊断算法研究[J]. 吉林化工学院学报, 2024, 41(3): 37-41.
[6] 王 影, 王 晨, 贾永涛, 刘 麒. 基于改进YOLOv5s的仓储货物检测算法研究[J]. 吉林化工学院学报, 2024, 41(1): 51-58.
[7] 董如意, 崔冉. 基于YOLOV5L的交通信号灯识别研究 [J]. 吉林化工学院学报, 2023, 40(9): 37-42.
[8] 刘麒, 盛德庆, 孙万龙, 王影.

基于改进YOLOv5s的水果目标检测研究 [J]. 吉林化工学院学报, 2023, 40(7): 34-41.

[9] 张 鋆, 李温温. RRT algorithm; path planning; target offset; variable step size; cubic B-spline curve 基于改进YOLOv5s的玉米田间杂草检测方法 [J]. 吉林化工学院学报, 2023, 40(5): 26-33.
[10] 金何. 基于深度学习融合网络的交直流电网故障诊断方法研究 [J]. 吉林化工学院学报, 2023, 40(3): 93-98.
[11] 辛 鹏, 杨剀勋, 文孝强. 基于改进SE-CNN的风电机组故障诊断方法研究 [J]. 吉林化工学院学报, 2023, 40(1): 34-40.
[12] 余思黔, 赵麒荣, 林嘉晨, 贾雁飞, 陈广大. 基于深度学习的核桃外壳缺陷检测 [J]. 吉林化工学院学报, 2022, 39(9): 80-85.
[13] 刘麒, 尹港 , 王影, 叶泽 . 基于深度学习的水面漂浮物识别算法设计 [J]. 吉林化工学院学报, 2022, 39(7): 28-33.
[14] 辛瑞昊 , 王甜甜 , 苗冯博 , 董哲原 , 马占森 , 冯欣. 基于深度学习的机械轴承故障智能诊断 [J]. 吉林化工学院学报, 2022, 39(11): 25-29.
[15] 朱莉, 陈辉. 基于深度学习的单幅图像三维重建算法 [J]. 吉林化工学院学报, 2020, 37(1): 58-62.
[1] . [J]. Journal of Jilin Institute of Chemical Technology, 2018, 35(9): 0 .
[2] . [J]. Journal of Jilin Institute of Chemical Technology, 2018, 35(9): 0 .
[3] SHAO Bao-li, LU Da, ZHAO Dong-hui. The Application of Dimensional Analysis in the Physical Quantity Conversion between Physical System and Numerical System [J]. Journal of Jilin Institute of Chemical Technology, 2018, 35(9): 1 -3 .
[4] ZHANG Jian, ZHAO Xiang, QU Bo, WU Qi, LIU Yu-tong, LI Yu-shi, LIU Qun. Application of Phosphorus-sulfur-nitrogen Composite Flame Retardant in Cotton Fabric [J]. Journal of Jilin Institute of Chemical Technology, 2018, 35(9): 4 -7 .
[5] WU Ping, REN Hong, LU Fei, WEI Qingling. A Functional Material on Recognition of Zn(II) ions based on the New Azo Compound [J]. Journal of Jilin Institute of Chemical Technology, 2018, 35(9): 8 -10 .
[6] YANG Yan-jun, WANG Ya-hong, Yang Xiu-dong. Process Aptimization of Surfactant Assisted Extraction of Total Polyphenols from Kyllinga Brevifolia Rottb [J]. Journal of Jilin Institute of Chemical Technology, 2018, 35(9): 11 -15 .
[7] LIU Jin-lu, LEI Yong-ping, WANG xiao-lin, ZHONG fang-li. Study on the  Purification Method of Total Saponins fromFruit of Rosa Davuvrica Pall. and its  Purification Method [J]. Journal of Jilin Institute of Chemical Technology, 2018, 35(9): 16 -23 .
[8] SONG Jian-gang, ZHONG Fang-li, WANG Xiao-lin, LIN Yu. Study on Extraction of Anthocyanin from Aronia melanocarpa Fruit by Ionic liquid Ultrasound Assisted [J]. Journal of Jilin Institute of Chemical Technology, 2018, 35(9): 24 -31 .
[9] TAN Li-hui, TAN Hong-wu. The Crashworthiness Analysis of different Cross-Section Thin-Walled Components [J]. Journal of Jilin Institute of Chemical Technology, 2018, 35(9): 32 -35 .
[10] YU Wen-xin, ZHENG Kai, WANG Li-hui, LIU Hai-bo, Wang Jian-xin. The Influence of Magnetostrictive Transducer Radiation Plate material on Radiation Sound Field Distribution [J]. Journal of Jilin Institute of Chemical Technology, 2018, 35(9): 36 -40 .
Viewed
Full text


Abstract

Cited

  Shared   
  Discussed