[1]王 勇,刘献泓,刘 盛,等.基于融合残差门控重建的轴承缺陷检测算法研究[J].机械与电子,2026,44(07):53-60.
　WANG Yong,LIU Xianhong,LIU Sheng,et al.Research on Bearing Defect Detection Algorithm Based on Fused Residual Gated Reconstruction[J].Machinery & Electronics,2026,44(07):53-60.
点击复制

基于融合残差门控重建的轴承缺陷检测算法研究()
分享到：

《机械与电子》[ISSN:1001-2257/CN:52-1052/TH]

卷:
44
期数:
2026年07期
页码:
53-60
栏目:
智能检测
出版日期:
2026-07-25

文章信息/Info

Title:
Research on Bearing Defect Detection Algorithm Based on Fused Residual Gated Reconstruction
文章编号:
1001-2257(2026)07-0053-08
作者:
王 勇1; 刘献泓2; 刘 盛1; 周 丹1; 汤志伟2; 周 锋2; 3
1.江苏省特种设备安全监督检验研究院盐城分院,江苏 盐城 224051;
2.盐城工学院信息工程学院,江苏 盐城224051;
3.盐城工学院光电信息技术研究所,江苏 盐城 224051
Author(s):
WANG Yong1; LIU Xianhong2; LIU Sheng1; ZHOU Dan1; TANG Zhiwei2; ZHOU Feng2; 3
(1.Yancheng Branch,Special Equipment Safety Supervision Inspection Institute of Jiangsu Province,Yancheng 224051,China;
2.School of Information Engineering,Yancheng Institute of Technology,Yancheng 224051,China;
3.Institute of Optoelectronic Information Technology,Yancheng Institute of Technology,Yancheng 224051,China)
关键词:
轴承表面缺陷检测; YOLOv12; 残差门控空间通道重建模块; SIoU-CIoU 组合损失
Keywords:
bearing surface defect detection; YOLOv12; residual-gated spatial-channel reconstruction
分类号:
TH133.3;TP391.4
文献标志码:
A
摘要:
针对轴承表面缺陷检测中微小特征易被背景淹没、复杂形态定位不准,提出YOLOv12改进算法。引入残差门控空间通道重建模块,自适应调节融合强度,增强微小缺陷感知;加权组合SIoU 与CIoU作为回归损失,协同优化角度与形状,兼顾方向敏感与不规则定位。所提模型以2 546 035 参数量和6.1 GFLOPs计算量,实现了88.6%的mAP50和64.2%的mAP95,相比基准YOLOv12n分别提升2.6百分点和2.1百分点。在相近或更小的模型规模下,所提方法取得了优于YOLOv8s、YOLOv11s、YOLOv12s等small版本以及RT-DETR、Deformable-DETR的检测精度,同时保持了极高的计算效率。
Abstract:
To address the issues of small defect features being easily overwhelmed by background clutter and inaccurate localization of complex morphologies in bearing surface defect detection,an improved YOLOv12 algorithm is proposed.A residual-gated spatial-channel reconstruction module is introduced to adaptively adjust the fusion intensity,thereby enhancing the perception of small defects.The regression loss is formulated as a weighted combination of SIoU and CIoU,which synergistically optimizes angular and shape attributes while balancing directional sensitivity and irregular localization.The proposed modelachieves 88.6% mAP50 and 64.2% mAP95 with only 2 546 035 parameters and 6.1 GFLOPs,outperforming the baseline YOLOv12n by 2.6 percentage points and 2.1 percentage points,respectively.With a comparable or even smaller model scale,the proposed method obtains higher detection accuracy than the small versions of YOLOv8s,YOLOv11s,and YOLOv12s,as well as RT-DETR and Deformable DETR, while maintaining exceptionally high computational efficiency.

参考文献/References:

[1] 魏波,冯乃勤,刘翠芳.深度残差神经网络下滚动轴承非平稳故障诊断[J].机械设计与制造,2026(1):201-206.

[2] 程锦锋,方贵盛,高惠芳.表面缺陷检测的机器视觉技术研究进展[J].计算机应用研究,2023,40(4):967-977.
[3] 徐彦威,李军,董元方,等.YOLO系列目标检测算法综述[J].计算机科学与探索,2024,18(9):2221-2238.
[4] Tian Yunjie,Ye Qixiang,Doermann D.YOLOv12:attention-centric real-time object detectors[PP/OL]∥ V1.arXiv(2025-02-18)[2026-03-04] https:∥doi.org/10.48550/arXiv.2502.12524.
[5] 张程,郑大斌.基于机器视觉的轴承表面缺陷自动检测系统设计[J].现代制造技术与装备,2026,62(3):82-84.
[6] Zheng Zhaohui,Wang Ping,Liu Wei,et al.Distance-IoU loss:faster and better learning for bounding box regression[C]∥Proceedings of the AAAI Conference on Artificial Intelligence.Washington:AAAI Press,2020,34(7):12993-13000.
[7] Rezatofighi H,Tsoi N,Gwak J,et al.Generalized intersection over union:a metric and a loss for bounding box regression[C]∥2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition.New York:IEEE,2019:658-666.
[8] 郭渊,周俊.基于机器视觉的轴承缺陷检测研究进展[J].机电工程,2024,41(5):761-774.
[9] Hu Jie,Shen Li,Sun Gang.Squeeze-and-excitation networks[C]∥2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition.New York:IEEE,2018:7132-7141.
[10] Woo S,Park J,Lee J Y,et al.CBAM:convolutional block attention module[C]∥Computer Vision-ECCV 2018.Cham:Springer,2018:3-19.
[11] Li Jiafeng,Wen Ying,He Lianghua.SCConv:spatial and channel reconstruction convolution for feature redundancy[ C]∥2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition.New York: IEEE,2023:6153-6162.
[12] Viswanath S.Spatially adaptive regularization for image restoration[D].Bangalore:Indian Institute of Science,2022.
[13] Gevorgyan Z.SIoU loss:More powerful learning for bounding box regression[PP/OL].V1.arXiv(2022-05-25)[2026-03-04].https:∥doi.org/10.48550/arXiv.2205.12740.
[14] Zhang Yifan,Ren Weiqiang,Zhang Zhang,et al.Focal and efficient IOU loss for accurate bounding box regression[ J].Neurocomputing,2022,506:146-157.
[15] Zhao Yian,Lv Wenyu,Xu Shangliang,et al.DETRs beat YOLOs on real-time object detection[C]∥2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition.New York:IEEE,2024:16965-16974.
[16] Zhu Xizhou,Su Weijie,Lu Lewei,et al.Deformable DETR: deformable transformers for end-to-end object detection[C]∥Proceedings of the 9th International Conference on Learning Representations,2021.

备注/Memo

备注/Memo:
收稿日期:2026-04-30
基金项目:江苏省特检院资助项目(KJ(Y)202633);盐城市重点研发计划工业领域竞争资助项目(YCBG2024023);盐城市应用基础研究计划资助项目(YCBK2025007);江苏省研究生科研与实践创新计划资助项目(SJCX24_2153,SJCX25_2202)
作者简介:王 勇 (1968-),男,江苏盐城人,高级工程师,研究方向为压力容器缺陷检测、压力容器安全性能测试与风险评估等;刘献泓 (2001-),男,江西丰城人,硕士研究生,研究方向为目标检测与识别、机器视觉和人工智能应用等;周 锋 (1981-),男,江苏盐城人,教授,博士研究生导师,研究方向为深度学习、计算机视觉和人工智能应用等,通信作者,E-mail:zfycit@ycit.edu.cn。
更新日期/Last Update: 2026-08-27