Please wait a minute...
浙江大学学报(工学版)  2020, Vol. 54 Issue (10): 1906-1914    DOI: 10.3785/j.issn.1008-973X.2020.10.006
计算机技术     
基于ViBe的端到端铝带表面缺陷检测识别方法
叶刚1(),李毅波1,2,3,*(),马逐曦2,成杰1
1. 中南大学 轻合金研究院,湖南 长沙 410083
2. 中南大学 机电工程学院,湖南 长沙 410083
3. 中南大学 高性能复杂制造国家重点实验室,湖南 长沙 410083
End-to-end aluminum strip surface defects detection and recognition method based on ViBe
Gang YE1(),Yi-bo LI1,2,3,*(),Zhu-xi MA2,Jie CHENG1
1. Light Alloy Research Institute, Central South University, Changsha 410083, China
2. School of Mechanical and Electrical Engineering, Central South University, Changsha 410083, China
3. State Key Laboratory of High Performance Complex Manufacturing, Central South University, Changsha 410083, China
 全文: PDF(1825 KB)   HTML
摘要:

针对铝带表面缺陷高精度检测要求以及传统算法识别率不佳的问题,提出端到端的表面缺陷检测与识别方法.从铝带表面初始图像序列中快速计算出平均图像,视为无缺陷背景图像,用于初始化ViBe算法的背景模型.采用ViBe算法从当前图像中分割出缺陷区域,对缺陷区域二值图像进行中值滤波和形态学运算,以去除噪声点和修补边缘,实现缺陷区域的准确提取.利用当前图像实时更新ViBe背景模型,以增加对光照变化的适应能力.提取缺陷外接矩形区域图像,归一化后输入到训练好的卷积神经网络中进行识别分类,得到分类结果.实验结果表明,提出方法的缺陷检出率为93.02%,缺陷识别率为99.86%,具有较好的应用价值.

关键词: 铝带表面缺陷缺陷检测缺陷识别ViBe卷积神经网络(CNN)    
Abstract:

An end-to-end surface defects detection and recognition method was proposed to solve the problem of high-precision detection of aluminum strip surface defects and the poor recognition rate of traditional algorithms. The average image was quickly calculated from the initial image sequence of aluminum strip surface, which was regarded as defect-free background image and was used to initialize the background model of the ViBe algorithm. The ViBe algorithm was used to segment the defect region from the current image. Median filtering and morphological operation were performed on the binary image of defect region to remove noise points and repair edges in order to accurately extract the defect region. The current image was used to update the ViBe background model in real time in order to increase the adaptability of the algorithm to illumination changes. The image of external rectangular region of the defect was extracted, normalized, and input into the trained convolutional neural networks for recognition and classification. The classification result was obtained. The experimental results show that the proposed method has a defect detection rate of 93.02% and a defect recognition rate of 99.86%, which has good application value.

Key words: surface defects of aluminum strip    defects detection    defects recognition    ViBe    convolutional neural network (CNN)
收稿日期: 2019-12-11 出版日期: 2020-10-28
CLC:  TP 391  
基金资助: “广西特聘专家”专项经费资助项目;中南大学中央高校基本科研业务费专项资金资助项目(2019zzts950)
通讯作者: 李毅波     E-mail: yeg2020@163.com;yibo.li@csu.edu.cn
作者简介: 叶刚(1993—),男,硕士生,从事机器视觉的研究. orcid.org/0000-0001-5359-5042. E-mail: yeg2020@163.com
服务  
把本文推荐给朋友
加入引用管理器
E-mail Alert
作者相关文章  
叶刚
李毅波
马逐曦
成杰

引用本文:

叶刚,李毅波,马逐曦,成杰. 基于ViBe的端到端铝带表面缺陷检测识别方法[J]. 浙江大学学报(工学版), 2020, 54(10): 1906-1914.

Gang YE,Yi-bo LI,Zhu-xi MA,Jie CHENG. End-to-end aluminum strip surface defects detection and recognition method based on ViBe. Journal of ZheJiang University (Engineering Science), 2020, 54(10): 1906-1914.

链接本文:

http://www.zjujournals.com/eng/CN/10.3785/j.issn.1008-973X.2020.10.006        http://www.zjujournals.com/eng/CN/Y2020/V54/I10/1906

图 1  ViBe背景点判别模型[17]
图 2  缺陷检测与识别算法流程图
图 3  基于VGG16的缺陷识别模型结构
图 4  图像采集平台
图 5  缺陷数据集样本分布
类别 黑点 凹凸包 凹坑 脏污 划痕
缺陷原图
分割结果
缺陷轮廓
缺陷外接矩形
表 1  缺陷检测结果示例
图 6  不同算法的分割效果对比
图 7  CNN训练期间的精度曲线和损失曲线
缺陷类别 黑点 凹凸包 凹坑 脏污 划痕 Pacc / %
黑点 98 0 0 1 0 98.99
凹凸包 0 212 0 0 0 100
凹坑 0 0 182 0 0 100
脏污 0 0 0 92 0 100
划痕 0 0 0 0 114 100
表 2  缺陷分类实验结果的混淆矩阵
1 康军伟, 张建辉, 曾宏凯 国内外铝加工行业状况浅析[J]. 有色金属加工, 2017, 46 (6): 1- 5
KANG Jun-wei, ZHANG Jian-hui, ZENG Hong-kai Analysis of aluminum processing industry at home and abroad[J]. Nonferrous Metals Processing, 2017, 46 (6): 1- 5
2 陈良 国内铝板带箔加工行业的现状分析[J]. 有色金属加工, 2014, 43 (3): 1- 4
CHEN Liang Analysis of current situation of domestic aluminum plate, strip and foil processing industry[J]. Nonferrous Metals Processing, 2014, 43 (3): 1- 4
3 颜云辉. 机器视觉检测与板带钢质量评价[M]. 北京: 科学出版社, 2016.
4 陈凯华. 基于机器视觉的板材表面缺陷检测与识别算法研究[D]. 南昌: 华东交通大学, 2012.
CHEN Kai-hua. Research based on machine vision of plate surface defects detection and recognition algorithm [D]. Nanchang: East China Jiaotong University, 2012.
5 秦钟伟, 陈捷, 洪荣晶, 等 摩擦片表面缺陷的视觉显著性检测算法[J]. 浙江大学学报: 工学版, 2019, 53 (10): 1883- 1891
QIN Zhong-wei, CHEN Jie, HONG Rong-jing, et al Visual salience detection algorithm for surface defects of friction sheets[J]. Journal of Zhejiang University: Engineering Science, 2019, 53 (10): 1883- 1891
6 ZHAO J, YANG Y M, LI G. The cold rolling strip surface defect on-line inspection system based on machine vision [C]// 2010 2nd Pacific-Asia Conference on Circuits, Communications and System. Beijing: IEEE, 2010: 402-405.
7 YANG Y X, LI Q, CHEN P, et al. Strip surface defect detection algorithm based on background difference [C]//2010 2nd Pacific-Asia Conference on Circuits, Communications and System. Beijing: IEEE, 2010: 23-26.
8 YANG Y X, DENG Y, LI Q, et al. Strip surface defect recognition algorithm based on PCA and improved BP neural network [C]// 2010 2nd Pacific-Asia Conference on Circuits, Communications and System. Beijing: IEEE, 2010: 19-22.
9 SUN X H, GU J N, TANG S X, et al Research progress of visual inspection technology of steel products: a review[J]. Applied Sciences, 2018, 8 (11): 2195
doi: 10.3390/app8112195
10 BENGIO Y, COURVILLE A, VINCENT P Representation learning: a review and new perspectives[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2013, 35 (8): 1798- 1828
doi: 10.1109/TPAMI.2013.50
11 LECUN Y, BENGIO Y, HINTON G Deep learning[J]. Nature, 2015, 521 (7553): 436
doi: 10.1038/nature14539
12 LI J Y, SU Z F, GENG J H, et al. Real-time detection of steel strip surface defects based on improved YOLO detection network [C/OL]// 5th IFAC Workshop on Mining, Mineral and Metal Processing. Shanghai: IFAC, 2018 [2019-12-01]. https://doi.org/ 10.1016/j.ifacol.2018.09.412.
13 REDMON J, DIVVALA S, GIRSHICK R, et al. You only look once: unified, real-time object detection [C]// 2016 IEEE Conference on Computer Vision and Pattern Recognition. Las Vegas: IEEE, 2016: 779-788.
14 TAO X, ZHANG D P, MA W Z, et al Automatic metallic surface defect detection and recognition with convolutional neural networks[J]. Applied Sciences, 2018, 8 (9): 1575
doi: 10.3390/app8091575
15 YI L, LI G Y, JIANG M 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
16 BARNICH O, DROOGENBROECK V M. VIBE: a powerful random technique to estimate the background in video sequences [C]// 2009 IEEE International Conference on Acoustics, Speech and Signal Processing. Taipei: IEEE, 2009: 945-948.
17 BARNICH O, DROOGENBROECK V M ViBe: a universal background subtraction algorithm for video sequences[J]. IEEE Transactions on Image Processing, 2011, 20 (6): 1709- 1724
doi: 10.1109/TIP.2010.2101613
18 陈亮, 陈晓竹, 范振涛 基于Vibe的鬼影抑制算法[J]. 中国计量学院学报, 2013, 24 (4): 425- 429
CHEN Liang, CHEN Xiao-zhu, FAN Zhen-tao Ghost suppression algorithm based on Vibe[J]. Journal of China University of Metrology, 2013, 24 (4): 425- 429
19 徐久强, 江萍萍, 朱宏博, 等 面向运动目标检测的ViBe算法改进[J]. 东北大学学报: 自然科学版, 2015, 36 (9): 1227- 1231
XU Jiu-qiang, JIANG Ping-ping, ZHU Hong-bo, et al An improved ViBe algorithm for moving object detection[J]. Journal of Northeastern University: Natural Science, 2015, 36 (9): 1227- 1231
20 BOUVRIE J. Notes on convolutional neural networks [EB/OL]. (2006-11-22) [2019-12-01]. http://web.mit.edu/jvb/www/papers/ cnn_tutorial.pdf.
21 张顺, 龚怡宏, 王进军 深度卷积神经网络的发展及其在计算机视觉领域的应用[J]. 计算机学报, 2019, 42 (3): 453- 482
ZHANG Shun, GONG Yi-hong, WANG Jin-jun The development of deep convolution neural network and its application on computer vision[J]. Chinese Journal of Computers, 2019, 42 (3): 453- 482
22 SIMONYAN K, ZISSERMAN A. Very deep convolutional networks for large-scale image recognition [EB/OL]. (2015-04-10) [2019-12-01]. http://arxiv.org/abs/1409.1556.
23 安宗权, 王匀 一种非线性扩散与图像差分的金属表面缺陷检测方法[J]. 表面技术, 2018, 47 (6): 277- 283
AN Zong-quan, WANG Yun A metal surface defect detection method based on nonlinear diffusion and image difference[J]. Surface Technology, 2018, 47 (6): 277- 283
[1] 陈杨波,伊国栋,张树有. 基于点云特征对比的曲面翘曲变形检测方法[J]. 浙江大学学报(工学版), 2021, 55(1): 81-88.
[2] 陈巧红,陈翊,李文书,贾宇波. 多尺度SE-Xception服装图像分类[J]. 浙江大学学报(工学版), 2020, 54(9): 1727-1735.
[3] 张彦楠,黄小红,马严,丛群. 基于深度学习的录音文本分类方法[J]. 浙江大学学报(工学版), 2020, 54(7): 1264-1271.
[4] 杨萍,王丹,康子健,李童,付利华,余悦任. 基于模式识别和集成CNN-LSTM的阵发性房颤预测模型[J]. 浙江大学学报(工学版), 2020, 54(5): 1039-1048.
[5] 李红光,郭英,眭萍,齐子森. 基于时频特征的卷积神经网络跳频调制识别[J]. 浙江大学学报(工学版), 2020, 54(10): 1945-1954.
[6] 贾子钰,林友芳,张宏钧,王晶. 基于深度卷积神经网络的睡眠分期模型[J]. 浙江大学学报(工学版), 2020, 54(10): 1899-1905.
[7] 吕艳,张萌,姜吴昊,倪益华,钱小鸿. 采用卷积神经网络的老年人跌倒检测系统设计[J]. 浙江大学学报(工学版), 2019, 53(6): 1130-1138.
[8] 赫贵然,李奇,冯华君,徐之海,陈跃庭. 基于CNN特征提取的双焦相机连续数字变焦[J]. 浙江大学学报(工学版), 2019, 53(6): 1182-1189.
[9] 董月,冯华君,徐之海,陈跃庭,李奇. Attention Res-Unet: 一种高效阴影检测算法[J]. 浙江大学学报(工学版), 2019, 53(2): 373-381.
[10] 王敬昌,陈岭,余珊珊,蒋晨书,吴勇. 基于门控循环单元的多因素感知短期游客人数预测模型[J]. 浙江大学学报(工学版), 2019, 53(12): 2357-2364.
[11] 秦钟伟,陈捷,洪荣晶,吴伟伟. 摩擦片表面缺陷的视觉显著性检测算法[J]. 浙江大学学报(工学版), 2019, 53(10): 1883-1891.
[12] 李浩,赵文杰,韩波. 基于滤波器裁剪的卷积神经网络加速算法[J]. 浙江大学学报(工学版), 2019, 53(10): 1994-2002.
[13] 张林, 程华, 房一泉. 基于卷积神经网络的链接表示及预测方法[J]. 浙江大学学报(工学版), 2018, 52(3): 552-559.
[14] 王卫星, 孙守迁, 李超, 唐智川. 基于卷积神经网络的脑电信号上肢运动意图识别[J]. 浙江大学学报(工学版), 2017, 51(7): 1381-1389.
[15] 唐有宝, 卜巍, 邬向前. 多层次MSER自然场景文本检测[J]. 浙江大学学报(工学版), 2016, 50(6): 1134-1140.