融合多层级特征的弱监督钢板表面缺陷检测算法

doi:10.12068/j.issn.1005-3026.2021.05.012

东北大学学报（自然科学版） ›› 2021, Vol. 42 ›› Issue (5): 687-692.DOI: 10.12068/j.issn.1005-3026.2021.05.012

融合多层级特征的弱监督钢板表面缺陷检测算法

何彧，宋克臣，张德富，颜云辉

(东北大学机械工程与自动化学院，辽宁沈阳110819)

修回日期:2020-11-24 接受日期:2020-11-24 发布日期:2021-05-20
通讯作者: 何彧
作者简介:何彧(1989-)，男，辽宁鞍山人，东北大学博士研究生; 颜云辉(1960-)，男，江苏丹阳人，东北大学教授，博士生导师.
基金资助:
基金项目；(半空) 基金项目.国家重点研发计划项目(2017YFB0304200)；国家自然科学基金资助项目(51805078)；中央高校基本科研业务费专项资金资助项目(N2003021).

Weakly-Supervised Steel Plate Surface Defect Detection Algorithm by Integrating Multiple Level Features

HE Yu， SONG Ke-chen， ZHANG De-fu， YAN Yun-hui

School of Mechanical Engineering & Automation， Northeastern University， Shenyang 110819， China.

Revised:2020-11-24 Accepted:2020-11-24 Published:2021-05-20
Contact: YAN Yun-hui
About author:-
Supported by:
-

摘要/Abstract

摘要： 由于缺少实例级标签，使得深度神经网络在工业表面检测领域的应用受到了限制.为解决这一问题，本文面向实际的热轧钢板表面缺陷检测任务，提出基于弱监督学习的缺陷检测网络，该网络引入类激活映射模型，使用容易获取的图像级标签进行模型训练，进行钢板表面的缺陷检测.为了进一步提升检测精度和克服类激活映射模型原有的缺点，本文采用性能更优的残差网络作为主干网络进行特征提取，并提出了多层级特征融合网络进行类激活图的生成，来获取更多的细节信息和更准确的目标激活区域.通过在公开缺陷数据集NEU-CLS上进行实验，结果表明本文提出的方法能够在标签不完备的情况下进行缺陷检测任务，并取得0.68%分类错误率和17.75%定位错误率，胜过其他同类的方法.

关键词: 热轧钢板表面;缺陷检测;弱监督学习;多层级特征融合;类激活映射

Abstract: Due to the lack of instance-level annotations， the application of deep neural network in the field of industrial surface detection is limited. In order to solve this problem， a weakly-supervised-learning-based defect detection network is proposed for the practical defect detection task of hot rolled steel plate surfaces. This network introduces the class activation mapping model， which can be used to train the model with the image-level annotations that can be obtained relatively easily， and performs the defect detection on the surface of steel plates. In order to further improve the detection accuracy and overcome the shortcomings of the class activation mapping model， the residual network with better performances is used as the backbone network for feature extraction， and the multi-level feature integration network is proposed to generate the class activation maps. In this way， the network can obtain more detail information and activate target areas more accurately. Extensive experiments have been carried out on the NEU-CLS dataset， and the results show that the proposed method can detect defects with incomplete labels， and obtain the classification error rate of 0.68% and the localization error rate of 17.75%， which are better than the other related methods.

Key words: hot-rolled plate surface; defect detection; weakly-supervised learning; multi-level feature integration; class activation mapping

中图分类号:

TP391.4

何彧，宋克臣，张德富，颜云辉. 融合多层级特征的弱监督钢板表面缺陷检测算法[J]. 东北大学学报（自然科学版）, 2021, 42(5): 687-692.

HE Yu， SONG Ke-chen， ZHANG De-fu， YAN Yun-hui. Weakly-Supervised Steel Plate Surface Defect Detection Algorithm by Integrating Multiple Level Features[J]. Journal of Northeastern University(Natural Science), 2021, 42(5): 687-692.

参考文献

[1]Liu K，Wang H，Chen H，et al.Steel surface defect detection using a new haar-weibull-variance model in unsupervised manner［J］.IEEE Transactions on Instrumentation and Measurement，2017，66(10):2585-2596.
[2]Liu K，Luo N，Li A，et al.A new self-reference image decomposition algorithm for strip steel surface defect detection［J］.IEEE Transactions on Instrumentation and Measurement，2020，69(7):4732-4741.
[3]Dong Y，Tao D，Li X，et al.Texture classification and retrieval using shearlets and linear regression［J］.IEEE Transactions on Cybernetics，2015，45(3):358-369.
[4]Krizhevsky A，Sutskever I，Hinton G.ImageNet classification with deep convolutional neural networks［C］//Neural Information Processing Systems(NIPS).New York，2012:1097-1105.
[5]Ren S，He K，Girshick R，et al.Faster R-CNN:towards real-time object detection with region proposal networks［C］//Neural Information Processing Systems(NIPS).Montreal，2015:91-99.
[6]Zhou S，Chen Y，Zhang D，et al.Classification of surface defects on steel sheet using convolutional neural networks［J］.Materiali in Tehnologije，2017，51(1):123-131.
[7]Yi L，Li G，Jiang M.An end to end steel strip surface defects recognition system based on convolutional neural networks［J］.Steel Research International，2017，88(2):176-187.
[8]He Y，Song K，Meng Q G，et al.An end-to-end steel surface defect detection approach via fusing multiple hierarchical features.［J］.IEEE Transactions on Instrumentation and Measurement，2020，69(4):1493-1504.
[9]Zhou B，Khosla A，Lapedriza A，et al.Learning deep features for discriminative localization［C］//Computer Vision and Pattern Recognition(CVPR).Las Vegas，2016:2921-2929.
[10]He K，Zhang X，Ren S，et al.Deep residual learning for image recognition［C］//Computer Vision and Pattern Recognition(CVPR).Boston，2015:770-778.
[11]Zeiler M D，Fergus R.Visualizing and understanding convolutional networks［C］//European Conference on Computer Vision(ECCV).Zurrch，2014:818-833.
[12]Yosinski J，Clune J，Bengio Y，et al.How transferable are features in deep neural networks［C］//Neural Information Processing Systems(NIPS).Montreal，2014:3320-3328.
[13]Goldman S A，Zhou Y.Enhancing supervised learning with unlabeled data［C］//International Conference on Machine Learning(ICML).Stanford，2000:327-334.
[14]Song K，Yan Y.A noise robust method based on completed local binary patterns for hot-rolled steel strip surface defects［J］.Applied Surface Science:Part B，2013，285:858-864.
[15]Lecun Y，Bottou L.Gradient-based learning applied to document recognition［J］.Proceedings of the IEEE，1998，86(11):2278-2324.

融合多层级特征的弱监督钢板表面缺陷检测算法

Weakly-Supervised Steel Plate Surface Defect Detection Algorithm by Integrating Multiple Level Features

RichHTML

PDF (PC)

可视化

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 15

编辑推荐

Metrics

本文评价

[1]	丁其川，王力，刘成. 融合长距离信道注意力与病理特征的肺结节分类[J]. 东北大学学报（自然科学版）, 2023, 44(4): 476-485.
[2]	杨丹，刘国如，任梦成，裴宏杨. 多尺度卷积核U-Net模型的视网膜血管分割方法[J]. 东北大学学报（自然科学版）, 2021, 42(1): 7-14.
[3]	原培新，陈鼎夫. 双能X射线高动态范围安检图像压缩算法[J]. 东北大学学报（自然科学版）, 2021, 42(1): 96-101.
[4]	张田，田勇，王子，王昭东. 基于清晰度评价的自适应阈值图像分割法[J]. 东北大学学报:自然科学版, 2020, 41(9): 1231-1238.
[5]	徐久强，张金鹏，贾玉其，邵建新. 基于集成学习的束支传导阻滞识别方法[J]. 东北大学学报:自然科学版, 2020, 41(3): 321-326.
[6]	徐礼胜，宋代远，刘文彦，王瑜璠. 用不同外周动脉波形重建主动脉波形的对比研究[J]. 东北大学学报:自然科学版, 2020, 41(2): 188-192.
[7]	冯宝，李昌林，李智，刘壮盛. 基于活动轮廓模型和影像组学的乳腺癌LVI状态预测[J]. 东北大学学报:自然科学版, 2020, 41(2): 193-199.
[8]	石春鹤，吴成东. 基于加权稀疏非负矩阵分解的车脸识别算法[J]. 东北大学学报:自然科学版, 2019, 40(10): 1376-1381.
[9]	张云洲，郑瑞，暴吉宁，朱尚栋. 基于多个相关滤波器的行人跟踪尺度算法[J]. 东北大学学报:自然科学版, 2019, 40(9): 1228-1234.
[10]	陈红，于晓升，吴成东，孙鹏. 参数化水平集活动轮廓模型的快速图像分割算法[J]. 东北大学学报:自然科学版, 2019, 40(1): 6-10.
[11]	张良钰，边天元，徐礼胜. 冠脉狭窄对外周与中心动脉间传递函数的影响[J]. 东北大学学报:自然科学版, 2018, 39(12): 1702-1707.
[12]	赵海，李雄峰，朱宏博，王彬. 基于分形维数特征的肺结节形状建模[J]. 东北大学学报:自然科学版, 2018, 39(11): 1545-1551.
[13]	窦圣昶，赵海，朱宏博，王彬. 修正与融合的肺部分割算法[J]. 东北大学学报:自然科学版, 2018, 39(11): 1562-1566.
[14]	徐礼胜，张书琪，牛潇，徐阳. 基于全卷积网络的左心室射血分数自动检测[J]. 东北大学学报:自然科学版, 2018, 39(11): 1572-1576.
[15]	齐林，吕旭阳，杨本强，徐礼胜. 基于全卷积网络迁移学习的左心室内膜分割[J]. 东北大学学报:自然科学版, 2018, 39(11): 1577-1582.