[1]胡峻峰,张志超,赵亚凤*,等. 基于迁移学习的竹片缺陷识别[J].西北林学院学报,2021,36(5):190-196.[doi:10.3969/j.issn.1001-7461.2021.05.29]
　HU Jun-feng,ZHANG Zhi-chao,ZHAO Ya-feng*,et al. Defect Recognition of Bamboo Sheets Based on Transfer Learning[J].JOURNAL OF NORTHWEST FORESTRY UNIVERSITY,2021,36(5):190-196.[doi:10.3969/j.issn.1001-7461.2021.05.29]
点击复制

 基于迁移学习的竹片缺陷识别()
分享到：

《西北林学院学报》[ISSN:1001-7461/CN:61-1202/S]

卷:
第36卷
期数:
2021年第5期
页码:
190-196
栏目:
出版日期:
2021-09-30

文章信息/Info

Title:
 Defect Recognition of Bamboo Sheets Based on Transfer Learning
文章编号:
1001-7461(2021)05-0190-07
作者:
 胡峻峰1; 张志超1; 赵亚凤2*; 陈振2
 (1.东北林业大学 机电工程学院，黑龙江 哈尔滨 150040；2.东北林业大学 信息与计算机工程学院，黑龙江 哈尔滨 150040)
Author(s):
 HU Jun-feng1; ZHANG Zhi-chao1; ZHAO Ya-feng2*; CHEN Zhen2
 (1.College of Mechanical and Electrical Engineering,Northeast Forestry University,Harbin 150040,Heilongjiang,China; 2.College of Information and Computer Engineering,Northeast Forestry University,Harbin 150040,Heilongjiang,China)
关键词:
 ; 卷积神经网络; 深度学习; 迁移学习; 混淆矩阵
Keywords:
 ; convolutional neural network;  deep learning;  transfer learning;  confusion matrix
分类号:
TP391.4
DOI:
10.3969/j.issn.1001-7461.2021.05.29
文献标志码:
A
摘要:
 使用竹片图像实现竹片缺陷自动识别，目前深度学习可以有效地解决该类问题，但是必须使用大量样本数据做训练才能获得较高的识别准确率。当图像数量有限时，利用基于迁移学习的方法，把经过预训练的卷积神经网络模型进行迁移，即共享卷积层和池化层的权重参数，调整新网络模型的超参数，并建立一个包含4种共计6 360张竹片缺陷图像的数据库，把图片分成4种训练集测试集形式，即80%训练、20%测试；60%训练、40%测试；40%训练、60%测试；20%训练、80%测试，分别利用支持向量机SVM分类方法、深度学习方法和迁移学习方法进行训练和测试，并将这3种方法作对比。最后，通过构建竹片缺陷识别的混淆矩阵对迁移学习进行具体分析与说明。结果表明，按照80%训练、20%测试的识别准确率最高，通过迁移学习得到的竹片缺陷最高识别精度分别达到98.97%，比普通深度学习提高了11.55% ，比SVM分类方法提高了13.04%。说明迁移学习比普通深度学习和传统支持向量机SVM分类方法更适合用于小样本数据集的分类识别，并且效果优于普通深度学习和 SVM 分类方法。
Abstract:
 At present,deep learning can effectively realize the automatic identification of the defects of bamboo sheets,in which however,a large number of sample data must be used for training in order to achieve high recognition accuracy.To solve the problem of limited images,we tried to use a new method that was based on the migration of learning to achieve the defect identification of bamboo sheets.In this method,a the migration operation was conducted for the preliminarily trained convolution neural network model,namely,to share the weight of convolution and pooling layer parameters,and to adjust the new super network model parameters.A database that included four types of defected bamboo sheet images with the number of 6 360 were established.The pictures were divided into four training and testing sets,namely,80% (training) & 20% (testing),60%&40%,40%&60%,20%&80%,the four types were trained and tested by three methods:support vector machine SVM classification method,the deep learning method,and migration study methods.The results by three methods were compared.Finally,the transfer learning was analyzed and explained concretely by constructing the confusion matrix of bamboo sheet defect recognition.The experimental results showed that 80% training &20% testing had the highest recognition accuracy.The highest recognition accuracy of bamboo sheet defects obtained by transfer learning was 98.97%,11.55% higher than the deep learning method and 13.04% higher than the SVM classification method.It showed that the transfer learning was more suitable for the classification and recognition of small sample data sets than the common deep learning and the traditional support vector machine SVM classification methods,and the results were better than the common deep learning and SVM classification methods.

相似文献/References:

[1]李亚藏,梁彦兰,王庆成. 铅对山梨和山荆子光合作用和叶绿素荧光特性的影响[J].西北林学院学报,2012,27(05):21.
　LI Ya-cang,LIANG Yan-lan,WANG Qing-cheng. Influence of Pb on Photosynthesis and Chlorophyll Fluorescence Characteristics in Pyrus ussuriensis and Malus baccata [J].JOURNAL OF NORTHWEST FORESTRY UNIVERSITY,2012,27(5):21.
[2]张小咏,李佳,杨艳昭,等. 基于SWAT模型的长江源区径流模拟[J].西北林学院学报,2012,27(05):38.
　ZHANG Xiao-yong,LI Jia,YANG Yan-zhao,et al. Runoff Simulation of the Catchment of the Headwaters of the Yangtze River Based on SWAT Model[J].JOURNAL OF NORTHWEST FORESTRY UNIVERSITY,2012,27(5):38.
[3]杨红旗,陈广辉,王金林. PVAc改性酚醛树脂制备铝木复合材料研究[J].西北林学院学报,2013,28(01):170.[doi:10.3969/j.issn.1001-7461.2013.01.34]
　YANG Hong-qi,CHEN Guang-hui,WANG Jin-lin. PF Adhesive Modified by PVAc for Manufacturing Technology of Decorative Fiberboard with Aluminum Foils[J].JOURNAL OF NORTHWEST FORESTRY UNIVERSITY,2013,28(5):170.[doi:10.3969/j.issn.1001-7461.2013.01.34]
[4]江涛,李秀荣,谢延军,等. 微波预处理对巨尾桉木材渗透性的影响[J].西北林学院学报,2013,28(01):174.[doi:10.3969/j.issn.1001-7461.2013.01.35]
　JIANG Tao,LI Xiu-rong,XIE Yan-jun,et al. Effects of Microwave Pretreatment on the Permeability of Eucalyptus grandis×Eucalyptus urophylla[J].JOURNAL OF NORTHWEST FORESTRY UNIVERSITY,2013,28(5):174.[doi:10.3969/j.issn.1001-7461.2013.01.35]
[5]牛晓霆,王逢瑚,曹新民. 明清家具匠师原木下料的工艺原则及措施[J].西北林学院学报,2013,28(01):178.[doi:10.3969/j.issn.1001-7461.2013.01.36]
　NIU Xiao-ting,WANG Feng-hu,CAO Xin-min. Techniques and Principles of Log Cuttings of Furniture Makers in Ming and Qing Dynasties[J].JOURNAL OF NORTHWEST FORESTRY UNIVERSITY,2013,28(5):178.[doi:10.3969/j.issn.1001-7461.2013.01.36]
[6]吕金阳,覃卓凯,舒辉,等. 银杉木材构造美学价值[J].西北林学院学报,2013,28(01):183.[doi:10.3969/j.issn.1001-7461.2013.01.37]
　LV Jin-yang,QIN Zhuo-kai,SHU Hui,et al. Esthetical Values in Wood Structure of Cathaya argyrophylla[J].JOURNAL OF NORTHWEST FORESTRY UNIVERSITY,2013,28(5):183.[doi:10.3969/j.issn.1001-7461.2013.01.37]
[7]杨艳红,张海敏,孙景荣. 基于传统文化视角的现代家具设计探析[J].西北林学院学报,2013,28(01):188.[doi:10.3969/j.issn.1001-7461.2013.01.38]
　YANG Yan-hong,ZHANG Hai-min,SUN Jing-rong. Study of Modern Furniture Design Based on Traditional Culture[J].JOURNAL OF NORTHWEST FORESTRY UNIVERSITY,2013,28(5):188.[doi:10.3969/j.issn.1001-7461.2013.01.38]
[8]黄圣游,叶喜. 新中式家具“中式”传统的界定[J].西北林学院学报,2013,28(01):192.[doi:10.3969/j.issn.1001-7461.2013.01.39]
　HUANG Sheng-You,YE Xi. Definition of Traditional Element in New Chinese Style Furniture[J].JOURNAL OF NORTHWEST FORESTRY UNIVERSITY,2013,28(5):192.[doi:10.3969/j.issn.1001-7461.2013.01.39]
[9]韩维生,赵明磊,王宏斌. 基于阈限理论的设计尺度体系[J].西北林学院学报,2013,28(01):197.[doi:10.3969/j.issn.1001-7461.2013.01.40]
　HAN Wei-sheng,ZHAO Ming-lei,WANG Hong-bin. Design Scale System Based on Threshold Theory[J].JOURNAL OF NORTHWEST FORESTRY UNIVERSITY,2013,28(5):197.[doi:10.3969/j.issn.1001-7461.2013.01.40]
[10]宋杰,侯建军,申黎明. 按摩椅按摩头包覆层对按摩舒适性的影响[J].西北林学院学报,2013,28(01):202.[doi:10.3969/j.issn.1001-7461.2013.01.41]
　SONG Jie,HOU Jian-jun,SHEN Li-ming. Influence of Massager Coat Layer of Massage Chairs on the Degree of Comfort[J].JOURNAL OF NORTHWEST FORESTRY UNIVERSITY,2013,28(5):202.[doi:10.3969/j.issn.1001-7461.2013.01.41]
[11]刘嘉政. 基于卷积神经网络的小样本树皮图像识别方法[J].西北林学院学报,2019,34(4):230.[doi:10.3969/j.issn.1001-7461.2019.04.35]
　LIU Jia-zheng. Small Sample Bark Image Recognition Method Based on Convolutional Neural Network[J].JOURNAL OF NORTHWEST FORESTRY UNIVERSITY,2019,34(5):230.[doi:10.3969/j.issn.1001-7461.2019.04.35]

备注/Memo

备注/Memo:
 收稿日期：2020-09-18修回日期：2020-10-12
基金项目：中央高校基本科研业务费专项资金(2572017CB10，2572019BF09)；黑龙江省博士后经费(LBH-Z16011)；黑龙江省自然科学基金项目(QC2015080)。
作者简介：胡峻峰，博士，副教授。研究方向：深度学习和机器视觉。E-mail：nefuhjf@126.com
*通信作者：赵亚凤，博士，副教授。研究方向：深度学习和机器视觉。E-mail：nefuzyf@126.com
更新日期/Last Update: