卷积神经网络在印刷品缺陷检测的应用
|
摘要
目的为了改善传统机器检测印刷产品缺陷存在误费率高的不足。方法提出以卷积神经网络为控制核心的印刷品缺陷检测系统。设计可在实际检测中应用的卷积神经网络,设计在线印刷质量检测系统的硬件结构。结果对结构相同而训练次数、学习率不同的卷积神经网络进行了缺陷检测的性能对比,验证了该卷积神经网络在学习率小于0.01时,可以获得较好的识别效果;在学习率大于0.05时,网络不容易收敛。网络训练次数越多,精度越高,相应的训练时间也较长。在满足快速性和精确度的条件下,确定了适应某印刷品的缺陷检验网络训练次数为50,学习率为0.005,此时的识别率为90%。结论经过实验证明,该检测系统具有良好的缺陷识别能力,缺陷类型的分类准确率较高。该系统具有一定的实用价值。
The paper aims to overcome the shortcoming of high error rate in traditional machine inspection of printing defects. A printing defect detection system based on convolutional neural network was proposed. The convolution neural network which can be used in practical inspection was designed, and the hardware structure of online printing quality detection system was designed. The defect detection performance of convolutional neural network with the same structure but different training times and different learning rate parameters was compared. It was verified that convolutional neural network with learning rate below 0.01 can achieve better recognition effect, while the learning rate network is hard to be converged when the learning rate is above 0.05. The more the number of network training, the higher the accuracy,and the longer the training time correspondingly. Under the premise of meeting the requirements on rapidity and accuracy,the number of network training for defect inspection of a printed matter was determined to be 50 times, and the learning rate was 0.005. In this case, the recognition rate was 90%. Experiments show that the detection system has good defect recognition ability and high classification accuracy of defect types. The system has certain practical value.
The paper aims to overcome the shortcoming of high error rate in traditional machine inspection of printing defects. A printing defect detection system based on convolutional neural network was proposed. The convolution neural network which can be used in practical inspection was designed, and the hardware structure of online printing quality detection system was designed. The defect detection performance of convolutional neural network with the same structure but different training times and different learning rate parameters was compared. It was verified that convolutional neural network with learning rate below 0.01 can achieve better recognition effect, while the learning rate network is hard to be converged when the learning rate is above 0.05. The more the number of network training, the higher the accuracy,and the longer the training time correspondingly. Under the premise of meeting the requirements on rapidity and accuracy,the number of network training for defect inspection of a printed matter was determined to be 50 times, and the learning rate was 0.005. In this case, the recognition rate was 90%. Experiments show that the detection system has good defect recognition ability and high classification accuracy of defect types. The system has certain practical value.
引文
[1]潘全玲,张蕾.印刷品在线检测系统在应用中出现的问题及解决方案[J].包装工程,2007,28(4):45-47.PAN Quan-ling,ZHANG Lei.Problem Solution of On-line Defect Detection System for Printed Matter[J].Packaging Engineering,2007,28(4):45-47.
[2]张慧,王坤峰,王飞跃.深度学习在目标视觉检测中的应用进展与展望[J].自动化学报,2017,43(8):1289-1305.ZHANG Hui,WANG Kun-feng,WANG Fei-yue.Advances and Perspectives on Applications of Deep Learning in Visual Object Detection[J].Acta Automatica Sinica,2017,43(8):1289-1305.
[3]常亮,邓小明,周明全,等.图像理解中的卷积神经网络[J].自动化学报,2016,42(9):1300-1312.CHANG Liang,DEND Xiao-ming,ZHOU Ming-quan,et al.Convolutional Neural Networks in Image Understand[J].Acta Automatica Sinica,2016,42(9):1300-1312.
[4]余文勇,周祖德,陈幼平.一种高速印刷品缺陷在线检测系统[J].华中科技大学学报,2006,34(6):80-83.YU Wen-yong,ZHOU Zu-de,CHEN You-ping.A High Speeded-inspection System of Defects in Printed Matter on Line[J].Journal of Huazhong University of Science and Technology(Nature Science),2006,34(6):80-83.
[5]杨祖彬,代小红.基于图像配准的食品包装印刷缺陷检测与实现[J].计算机科学,2015,42(8):319-322.YANG Zu-bin,DAI Xiao-hong.Printing Defects Detection and Realization in Food Packaging Based on Image Registration[J].Computer Science,2015,42(8):319-322.
[6]罗中兵,付忠良,阮波.一种改进的模板匹配算法[J].计算机应用,2002(3):28-30.LUO Zhong-bing,FU Zhong-liang,RUAN Bo.An Improved Template Matching Algorithm[J].Computer Applications,2002(3):28-30.
[7]王文举,赵萍,陈伟,等.彩色印刷品缺陷快速精确检测方法研究[J].包装工程,2015,36(17):112-118.WANG Wen-ju,ZHAO Ping,CHEN Wei,et al.A Fast and Accurate Method of Defect Detection of Color Printing Image[J].Packaging Engineering,201536(17):112-118.
[8]胡方尚,郭慧,邢金鹏,等.基于印刷缺陷检测的图像配准方法研究[J].光学技术,2017,43(1):19-24.HU Fang-shang,GUO Hui,XING Jin-peng,et al.Image Registration Based on Label Printing Defect Detection[J].Optical Technique,2017,43(1):19-24.
[9]胡方尚,郭慧.基于改进多类支持向量机的印刷缺陷检测[J].华东理工大学学报(自然科学版),2017,43(1):143-148.HU Fang-shang,GUO Hui.Printing Defects Inspection Based on Improved Multi-class Support Vector Machine[J].Journal of East China University of Science and Technology:Natural Science Edition,2017,43(1):143-148.
[10]王素芬.多维彩色图像印刷缺陷快速检测仿真[J].计算机仿真,2017,34(6):411-414.WANG Su-fen.Fast Detection of Multi Dimensional Color Image Printing Defects[J].Computer Simulation,2017,34(6):411-414.
[11]孙巍巍,孙权森,夏德深.一种烟标在线检测系统的设计与实现[J].微电子学与计算机,2009,26(10):107-110.SUN Wei-wei,SUN Quan-sen,XIA De-shen.A Design and Implementation of Online Defect Detection System for Cigarette Packet[J].Microelectronics and Computer,2009,26(10):107-110.
[12]初红艳,李鹏,蔡力钢.基于控制图和神经网络的印刷过程质量智能监控技术[J].北京工业大学学报,2012,38(3):340-344.CHU Hong-yan,LI Peng,CAI Li-gang.Intelligent Monitoring Technology of Printing Process Quality Based on Control Chart and Neural Network[J].Journal of Beijing Polytechnic University,2012,38(3):340-344.
[13]孙碧亮,尚会超,余文勇,等.一种用于高速印刷品质量检测的预处理算法[J].计算机工程与应用,2006,42(10):188-191.SUN Bi-liang,SHANG Hui-chao,YU Wen-yong,et al.A Pretreatment Algorithm for an Online Quality Detecting System in the High-speed Presswork Production Line[J].Computer Engineering and Applications,2006,42(10):188-191.
[14]穆向阳,张太镒.机器视觉系统的设计[J].西安石油大学学报(自然科学版),2007,22(6):104-109.MU Xiang-yang,ZHANG Tai-yi.Design of a Machine Vision System[J].Journal of Xi'an Shiyou University:Natural Science Edition,2007(6):104-109.
[15]石绘,余文勇.商业票证印刷缺陷检测方法的研究[J].武汉理工大学学报,2008,30(5):148-150.SHI Hui,YU Wen-yong.Research on the Way of On-line Quality Inspection of Commerce Note Printing[J].Journal of Wuhan University of Technology2008,30(5):148-150.
[16]王任华,沈忙作.自动对焦算法研究[J].光电工程2000(4):11-13.WANG Ren-hua,SHEN Mang-zuo.An Algorithm Study for Automatic Focusing[J].Opto-Electronic Engineering,2000(4):11-13.
[2]张慧,王坤峰,王飞跃.深度学习在目标视觉检测中的应用进展与展望[J].自动化学报,2017,43(8):1289-1305.ZHANG Hui,WANG Kun-feng,WANG Fei-yue.Advances and Perspectives on Applications of Deep Learning in Visual Object Detection[J].Acta Automatica Sinica,2017,43(8):1289-1305.
[3]常亮,邓小明,周明全,等.图像理解中的卷积神经网络[J].自动化学报,2016,42(9):1300-1312.CHANG Liang,DEND Xiao-ming,ZHOU Ming-quan,et al.Convolutional Neural Networks in Image Understand[J].Acta Automatica Sinica,2016,42(9):1300-1312.
[4]余文勇,周祖德,陈幼平.一种高速印刷品缺陷在线检测系统[J].华中科技大学学报,2006,34(6):80-83.YU Wen-yong,ZHOU Zu-de,CHEN You-ping.A High Speeded-inspection System of Defects in Printed Matter on Line[J].Journal of Huazhong University of Science and Technology(Nature Science),2006,34(6):80-83.
[5]杨祖彬,代小红.基于图像配准的食品包装印刷缺陷检测与实现[J].计算机科学,2015,42(8):319-322.YANG Zu-bin,DAI Xiao-hong.Printing Defects Detection and Realization in Food Packaging Based on Image Registration[J].Computer Science,2015,42(8):319-322.
[6]罗中兵,付忠良,阮波.一种改进的模板匹配算法[J].计算机应用,2002(3):28-30.LUO Zhong-bing,FU Zhong-liang,RUAN Bo.An Improved Template Matching Algorithm[J].Computer Applications,2002(3):28-30.
[7]王文举,赵萍,陈伟,等.彩色印刷品缺陷快速精确检测方法研究[J].包装工程,2015,36(17):112-118.WANG Wen-ju,ZHAO Ping,CHEN Wei,et al.A Fast and Accurate Method of Defect Detection of Color Printing Image[J].Packaging Engineering,201536(17):112-118.
[8]胡方尚,郭慧,邢金鹏,等.基于印刷缺陷检测的图像配准方法研究[J].光学技术,2017,43(1):19-24.HU Fang-shang,GUO Hui,XING Jin-peng,et al.Image Registration Based on Label Printing Defect Detection[J].Optical Technique,2017,43(1):19-24.
[9]胡方尚,郭慧.基于改进多类支持向量机的印刷缺陷检测[J].华东理工大学学报(自然科学版),2017,43(1):143-148.HU Fang-shang,GUO Hui.Printing Defects Inspection Based on Improved Multi-class Support Vector Machine[J].Journal of East China University of Science and Technology:Natural Science Edition,2017,43(1):143-148.
[10]王素芬.多维彩色图像印刷缺陷快速检测仿真[J].计算机仿真,2017,34(6):411-414.WANG Su-fen.Fast Detection of Multi Dimensional Color Image Printing Defects[J].Computer Simulation,2017,34(6):411-414.
[11]孙巍巍,孙权森,夏德深.一种烟标在线检测系统的设计与实现[J].微电子学与计算机,2009,26(10):107-110.SUN Wei-wei,SUN Quan-sen,XIA De-shen.A Design and Implementation of Online Defect Detection System for Cigarette Packet[J].Microelectronics and Computer,2009,26(10):107-110.
[12]初红艳,李鹏,蔡力钢.基于控制图和神经网络的印刷过程质量智能监控技术[J].北京工业大学学报,2012,38(3):340-344.CHU Hong-yan,LI Peng,CAI Li-gang.Intelligent Monitoring Technology of Printing Process Quality Based on Control Chart and Neural Network[J].Journal of Beijing Polytechnic University,2012,38(3):340-344.
[13]孙碧亮,尚会超,余文勇,等.一种用于高速印刷品质量检测的预处理算法[J].计算机工程与应用,2006,42(10):188-191.SUN Bi-liang,SHANG Hui-chao,YU Wen-yong,et al.A Pretreatment Algorithm for an Online Quality Detecting System in the High-speed Presswork Production Line[J].Computer Engineering and Applications,2006,42(10):188-191.
[14]穆向阳,张太镒.机器视觉系统的设计[J].西安石油大学学报(自然科学版),2007,22(6):104-109.MU Xiang-yang,ZHANG Tai-yi.Design of a Machine Vision System[J].Journal of Xi'an Shiyou University:Natural Science Edition,2007(6):104-109.
[15]石绘,余文勇.商业票证印刷缺陷检测方法的研究[J].武汉理工大学学报,2008,30(5):148-150.SHI Hui,YU Wen-yong.Research on the Way of On-line Quality Inspection of Commerce Note Printing[J].Journal of Wuhan University of Technology2008,30(5):148-150.
[16]王任华,沈忙作.自动对焦算法研究[J].光电工程2000(4):11-13.WANG Ren-hua,SHEN Mang-zuo.An Algorithm Study for Automatic Focusing[J].Opto-Electronic Engineering,2000(4):11-13.