基于机器视觉的强化木地板表面瑕疵检测方法研究
|
摘要
以小径材为原材料强化木地板俗称人造木地板,具有耐磨、美观、环保、防潮、阻燃、防蛀、安装便捷、易于清洁护理、经济实用等诸多优点,且具有价格优势,所以应用发展迅速。然而在国内强化木地板生产线上,对于产品表面粘贴的木纹装饰纸的外观质量检验目前还是依靠人工目测,劳动强度大,不仅容易造成视觉疲劳,而且由于不同检测员的认识不同,检测结果受主观因素影响产生差异,外观质量难以保证。同时人工检测速度慢,效率低。本课题试图探索人工智能方法的强化木地板表面质量检测,提高该生产线的自动化水平。
本文详细论述了利用视频摄像机进行强化木地板在线质量检测的具体方法,包括图像处理、特征提取和模式识别,这三个步骤是机器视觉研究的核心内容。作者首先通过调研北京科诺森华木地板厂等生产企业的技术要求,制定了课题研究的技术路线。在图像分割环节,本文试验了蚁群算法、最大类间方差算法和最大熵的遗传算法三种处理方法,结果表明遗传算法能够更好地提取地板表面浅色和深色缺陷,且运算速度较快。本文选取强化木地板表面缺陷的颜色特征和纹理特征进行了参数计算,并采用主成分分析法对这些参数做了降维处理,有效减少了后期运算的复杂程度。最后运用BP神经网络对实验样板进行了识别分类,完成了强化木地板表面质量智能检测方法研究的全过程。经MATLAB仿真表明,运用BP神经网络对强化木地板表面瑕疵进行检测效果不够理想,不能应用于实际生产。
Laminated flooring derived from smaller diameter of raw materials is also called artificial plank.It has many advantages, like wear-resisting, artistic, economical, practical, and resistant to damages by moisture, flame and moths, as well as convenient-installing.easy-nursing and environmental protecting, in addition, it has the advantage of lower cost.These characteristics make laminated flooring developing rapidly in China.However,in domestic laminated flooring production line,outward appearance quality inspection still depending on human visual. Due to the great labor intensity, it's easy to cause visual fatigue. What's worse, because of the various understanding of different testing personnel, it's difficult to unify the testing standards and guarantee the appearance quality.Thus, the test results are greatly affected by subjective factors, and the efficiency of artificial detection is low. This paper aims at using machine vision instead of artificial detection to improve the level of laminated floor producing automation line.
In this paper, the method of machine vision for on-line detection of laminated flooring is discussed, including image segmentation, feature extraction and classifier design. These three parts are the core contents of study. At the early stage of the research study, through visiting the factories of Kenuo Senhua floor factory of Beijing, we got a clear understanding of laminated flooring production, including the whole manufacture line. In the image segmentation, this paper puts forward three image segmentation methods: Image Segmentation by Ant Colony Algorithm; Image segmentation based on maximum between-cluster variance; Image Segmentation based on Genetic Algorithm and the maximum entropy. Through the comparison of the three methods, genetic algorithm is better and faster in segmenting floor surface defects. So it has used genetic algorithm for research in this paper. Research in laminated flooring surface defect feature, the characteristics of color and texture feature parameters are calculated, and the principal component analysis is used to reduce the complexity of the classifier operation.At last, it uses BP network structure to recognize and classfy the floor surface quality. By matlab simulation, it shows that using BP network to detect the laminated flooring defects is insufficient, so it cannot applied to practical production.
本文详细论述了利用视频摄像机进行强化木地板在线质量检测的具体方法,包括图像处理、特征提取和模式识别,这三个步骤是机器视觉研究的核心内容。作者首先通过调研北京科诺森华木地板厂等生产企业的技术要求,制定了课题研究的技术路线。在图像分割环节,本文试验了蚁群算法、最大类间方差算法和最大熵的遗传算法三种处理方法,结果表明遗传算法能够更好地提取地板表面浅色和深色缺陷,且运算速度较快。本文选取强化木地板表面缺陷的颜色特征和纹理特征进行了参数计算,并采用主成分分析法对这些参数做了降维处理,有效减少了后期运算的复杂程度。最后运用BP神经网络对实验样板进行了识别分类,完成了强化木地板表面质量智能检测方法研究的全过程。经MATLAB仿真表明,运用BP神经网络对强化木地板表面瑕疵进行检测效果不够理想,不能应用于实际生产。
Laminated flooring derived from smaller diameter of raw materials is also called artificial plank.It has many advantages, like wear-resisting, artistic, economical, practical, and resistant to damages by moisture, flame and moths, as well as convenient-installing.easy-nursing and environmental protecting, in addition, it has the advantage of lower cost.These characteristics make laminated flooring developing rapidly in China.However,in domestic laminated flooring production line,outward appearance quality inspection still depending on human visual. Due to the great labor intensity, it's easy to cause visual fatigue. What's worse, because of the various understanding of different testing personnel, it's difficult to unify the testing standards and guarantee the appearance quality.Thus, the test results are greatly affected by subjective factors, and the efficiency of artificial detection is low. This paper aims at using machine vision instead of artificial detection to improve the level of laminated floor producing automation line.
In this paper, the method of machine vision for on-line detection of laminated flooring is discussed, including image segmentation, feature extraction and classifier design. These three parts are the core contents of study. At the early stage of the research study, through visiting the factories of Kenuo Senhua floor factory of Beijing, we got a clear understanding of laminated flooring production, including the whole manufacture line. In the image segmentation, this paper puts forward three image segmentation methods: Image Segmentation by Ant Colony Algorithm; Image segmentation based on maximum between-cluster variance; Image Segmentation based on Genetic Algorithm and the maximum entropy. Through the comparison of the three methods, genetic algorithm is better and faster in segmenting floor surface defects. So it has used genetic algorithm for research in this paper. Research in laminated flooring surface defect feature, the characteristics of color and texture feature parameters are calculated, and the principal component analysis is used to reduce the complexity of the classifier operation.At last, it uses BP network structure to recognize and classfy the floor surface quality. By matlab simulation, it shows that using BP network to detect the laminated flooring defects is insufficient, so it cannot applied to practical production.
引文
1. 白雪冰,王克奇,王辉.基于灰度共生矩阵的木材纹理分类方法的研究[J].哈尔滨工业大学学报,2005,37(12)：1667-1670
2.曹国斌,刘雪蛟,王花.图像和机器视觉技术概述[J].电子工业专用设备,2008(8)
3.陈波.基于数学形态学的图像特征提取[J].科技情报开发与经济,2007,17(17)：161-162
4.杜奇,向健勇,袁胜春.一种改进的最大类间方差法[J].红外技术,2003,25(5)：33-36
5.高隽,人工神经网络原理及仿真实例[M].机械工业出版社,2003
6.GB／T 18102-2000,浸渍纸层压质地板[S]
7.高淑兰,刘英臣.强化复合地板的生产工艺及应用前景[J].林业科技.2001,26(2)：46-47
8.耿冠宏,孙伟,罗培.神经网络模式识别[J].软件导刊,2008,7(10)：81-83
9.耿瑞芳,曹辉.基于机器视觉的图像识别方法研究[J].自动化博览,2008(6)
10.关胜晓.机器视觉及其应用发展[J].自动化博览,2005,03：88-92
11.关子明,常文兵.基于主成分分析-RBF神经网络模型的备件预测研究[J].物流科技,2009(4)：122-126
12.郭德军,宋蛰存.基于灰度共生矩阵的纹理图像分类研究[J].林业机械与木工设备,2005,33(7)：21-23
13.韩彦芳,施鹏飞.基于蚁群算法的图像分割方法[J].计算机工程与应用,2004,40(18)：5-7
14.李红梅,宁功德.影响强化木地板质量因素的分析[J].林业机械与木工设备,2004,32(06)：33-34
15.李立源,龚坚,陈维甫.基于二维灰度直方图最佳一维投影的图像分割方法[J].电子学报,1996,22(3)：315-323
16.廖义奎,王志文.一种新的彩色图像特征提取方法及其应用[J].广西民族学院学报(自然科学版),2003.9(3)：48-52
17.牟洪波,基于人工神经网络的木材缺陷检测研究[D].东北林业大学,2006
18.彭兴璇.一种基于HSV颜色空间的图像检索方法[J].科技信息,2008.27：381-383
19.盛国芳,焦李成.基于遗传算法的最佳熵阈值的图像分割[J].计算机工程与应用,2003,12：103-105
20.王晗,白雪冰,王辉.基于空间灰度共生矩阵木材纹理分类识别的研究[J].森林工程,2007,23(1)：32-36
21.于立海,杨学春,徐凯宏.木材无损检测技术的研究现状与进展[J].森林工程,2001.17(6)：1-4
22.王业琴,赵志国.基于直方图和颜色矩方法的木材表面颜色特征的表达[J].林业科技,2006,31(5)：56-58
23.我国强化木地板行业现状以及发展趋势分析,www.ewood.cn,2008.3.5
24.我国强化木地板简述,www.ycjzw.cn,2008.1.2
25.肖建华.智能模式识别方法[M].华南理工大学出版社,2006
26.谢永华,基于分形理论木材表面缺陷识别的研究[D].东北林业大学.2006
27.徐琨,李燕.基于分块颜色矩和纹理特征的图像检索方法[J].西安石油大学学报(自然科学版),2005.20(2)：77-79
28.杨春梅,胡万义,白帆.木材缺陷检测理论及方法的发展[J].林业机械与木工设备,2004,32(3)：8-10
29.杨海峰,侯朝桢.基于二维灰度直方图的蚁群图像分割[J].激光与红外.2005,8：614-617
30.叶银兰.基于颜色和形状的图像检索[J].太原理工大学学报,2008.39(2)：127-131
31.易乔木,程金树等.一种基于机器视觉的玻璃质量在线检测系统[J].武汉理工大学学报,2007,29(5)：23-26
32.余成波.数字图像处理及MATLAB实现[M].重庆大学出版社,2003
33.张健,杨丽丽,鲍际平.基于蚁群算法的强化木地板表面缺陷图像处理技术研究[J].仪器仪表学报.2008,29(4)：109-112
34.张健,鲍际平等.基于遗传算法的强化木地板表面缺陷的图像分割[J].林业机械与木工设备,2008,36(7)：27-28
35.周开利.神经网络模型及其Mat]ab仿真程序设计[M].北京：清华大学出版社,2005
36.章毓晋.图像分割[M].北京：科学出版社,2001
37.朱贵锋.机器视觉在线检测图像处理技术研究[D].合肥工业大学,2006
38.朱志刚,林学訚等,数字图象处理[M].北京：电子工业出版社,2005
39. Bharati M H, LIU J J, MAC G, Image texture analysis: methods and comparisons [J]. Chemometrics & Intelligent Laboratory Systems,2004,72(1):57-65
40. Cao Z L, Hu J,Yang S Y. Object Recognition System Based on Image Texture Analysis. Journal of Tianjin Institute of Technology.2001:31-33.
41. Cheng H D, Chen Y H, Jiang X H. Threshold usingtwo2dimensional histogram and fuzzy entropy principle[J]. IEEE Transactions on Image Processing,2000,9(4):732-735
42. Cyra G, Tanaka C, Nakao T. On-line Control of Router Feed Speed U sing Acoustic Emission [J]. Forest Prod J.,1996,46(11/12):27-32
43. Faust T D. Real Time Measurement of Veneer Surface Rough-ness by Image Analysis [J]. Forest Prod.J.1987,37(6):34-40
44. Forrer J B, Funck J W. Dielectric Properties of Defects on Wood Surface [J]. Holzals Rohund Werkstoff.1998,56(1):25-29
45. Gonzalo A, Ruz, Pablo A, Estevez, Claudio A, Perez. A neurofuzzy color image segmentation method for wood surface defect detection [J]. Forest Products Journal,2006,60(4)
46. H.Elbehiery, A.Hefnawy, M.Elewa. Surface Defects Detection for Ceramic Tiles Using Image Processing and Morphological Techniques. Transactions on Engineering Computing and Technology,2005,5:158-162
47. I.J. Jolliffe. Principal Component Analysis [M]. Springer, New York,2002
48. Kenneth R, Castleman. Digital Image Processing[M].Publishing House of Electronic Industry,2005(7):430-432
49. K.Terada,N.Fujiwara, "Extraction of a smoke region using fractal coding", IEEE International Symposium on Communications and Information Technology,2004,ISCIT 2004,Volume 2,26-29 Oct.2004:659-662.
50. L emaster R, Beall C. The Use Optical Profilometer to Measure Surface Roughness in Mediu-m Density Fiberboard [J]. Forest Prod J.,1996,46(11/12):73-78
51. LawrenceK. Saul, SamT. Roweis, Think globally fit locally unsupervised learning of low dimensional manifolds [J]. Journal of Machine Learning Research 2003(4):179-155
52. Mark S. Nixon,Alberto S. Aguado. Feature Extraction and Image Processing [M]. Great Britain: Reed Educational and Professional Publishing Ltd,2006
53. MilanSonka,VaclavHlavac.图像处理、分析与机器视觉[M]2版.北京：人民邮电出版社,2002
54. Pckkarinen A. A Method for the Segmentation of Very High Spatial Resolution Images of Forested Landscapes[J].int.J.Remote Sensing,2002,23(14):2817-2836
55. Proceedings of the 15th International Wood Machining Seminar.2001:473-476
56. SONG J H. Maximum entropy thresholding image segmentation based on genetic algorithm [J].Electronic Engineer,2005,2(9):34-36
57. TAMURA H, MORI S, YAMAWAKI T. Texture features corresponding to visual perception [J]. IEEE Transactions on Systems, Man, and Cybernetics,2003,8(6):460-473
58. WANG X P, CAO L M. Genetic algorithms-theory, application and software implementation [M].Xi'an:Xi'an Jiao tong university press,2002
2.曹国斌,刘雪蛟,王花.图像和机器视觉技术概述[J].电子工业专用设备,2008(8)
3.陈波.基于数学形态学的图像特征提取[J].科技情报开发与经济,2007,17(17)：161-162
4.杜奇,向健勇,袁胜春.一种改进的最大类间方差法[J].红外技术,2003,25(5)：33-36
5.高隽,人工神经网络原理及仿真实例[M].机械工业出版社,2003
6.GB／T 18102-2000,浸渍纸层压质地板[S]
7.高淑兰,刘英臣.强化复合地板的生产工艺及应用前景[J].林业科技.2001,26(2)：46-47
8.耿冠宏,孙伟,罗培.神经网络模式识别[J].软件导刊,2008,7(10)：81-83
9.耿瑞芳,曹辉.基于机器视觉的图像识别方法研究[J].自动化博览,2008(6)
10.关胜晓.机器视觉及其应用发展[J].自动化博览,2005,03：88-92
11.关子明,常文兵.基于主成分分析-RBF神经网络模型的备件预测研究[J].物流科技,2009(4)：122-126
12.郭德军,宋蛰存.基于灰度共生矩阵的纹理图像分类研究[J].林业机械与木工设备,2005,33(7)：21-23
13.韩彦芳,施鹏飞.基于蚁群算法的图像分割方法[J].计算机工程与应用,2004,40(18)：5-7
14.李红梅,宁功德.影响强化木地板质量因素的分析[J].林业机械与木工设备,2004,32(06)：33-34
15.李立源,龚坚,陈维甫.基于二维灰度直方图最佳一维投影的图像分割方法[J].电子学报,1996,22(3)：315-323
16.廖义奎,王志文.一种新的彩色图像特征提取方法及其应用[J].广西民族学院学报(自然科学版),2003.9(3)：48-52
17.牟洪波,基于人工神经网络的木材缺陷检测研究[D].东北林业大学,2006
18.彭兴璇.一种基于HSV颜色空间的图像检索方法[J].科技信息,2008.27：381-383
19.盛国芳,焦李成.基于遗传算法的最佳熵阈值的图像分割[J].计算机工程与应用,2003,12：103-105
20.王晗,白雪冰,王辉.基于空间灰度共生矩阵木材纹理分类识别的研究[J].森林工程,2007,23(1)：32-36
21.于立海,杨学春,徐凯宏.木材无损检测技术的研究现状与进展[J].森林工程,2001.17(6)：1-4
22.王业琴,赵志国.基于直方图和颜色矩方法的木材表面颜色特征的表达[J].林业科技,2006,31(5)：56-58
23.我国强化木地板行业现状以及发展趋势分析,www.ewood.cn,2008.3.5
24.我国强化木地板简述,www.ycjzw.cn,2008.1.2
25.肖建华.智能模式识别方法[M].华南理工大学出版社,2006
26.谢永华,基于分形理论木材表面缺陷识别的研究[D].东北林业大学.2006
27.徐琨,李燕.基于分块颜色矩和纹理特征的图像检索方法[J].西安石油大学学报(自然科学版),2005.20(2)：77-79
28.杨春梅,胡万义,白帆.木材缺陷检测理论及方法的发展[J].林业机械与木工设备,2004,32(3)：8-10
29.杨海峰,侯朝桢.基于二维灰度直方图的蚁群图像分割[J].激光与红外.2005,8：614-617
30.叶银兰.基于颜色和形状的图像检索[J].太原理工大学学报,2008.39(2)：127-131
31.易乔木,程金树等.一种基于机器视觉的玻璃质量在线检测系统[J].武汉理工大学学报,2007,29(5)：23-26
32.余成波.数字图像处理及MATLAB实现[M].重庆大学出版社,2003
33.张健,杨丽丽,鲍际平.基于蚁群算法的强化木地板表面缺陷图像处理技术研究[J].仪器仪表学报.2008,29(4)：109-112
34.张健,鲍际平等.基于遗传算法的强化木地板表面缺陷的图像分割[J].林业机械与木工设备,2008,36(7)：27-28
35.周开利.神经网络模型及其Mat]ab仿真程序设计[M].北京：清华大学出版社,2005
36.章毓晋.图像分割[M].北京：科学出版社,2001
37.朱贵锋.机器视觉在线检测图像处理技术研究[D].合肥工业大学,2006
38.朱志刚,林学訚等,数字图象处理[M].北京：电子工业出版社,2005
39. Bharati M H, LIU J J, MAC G, Image texture analysis: methods and comparisons [J]. Chemometrics & Intelligent Laboratory Systems,2004,72(1):57-65
40. Cao Z L, Hu J,Yang S Y. Object Recognition System Based on Image Texture Analysis. Journal of Tianjin Institute of Technology.2001:31-33.
41. Cheng H D, Chen Y H, Jiang X H. Threshold usingtwo2dimensional histogram and fuzzy entropy principle[J]. IEEE Transactions on Image Processing,2000,9(4):732-735
42. Cyra G, Tanaka C, Nakao T. On-line Control of Router Feed Speed U sing Acoustic Emission [J]. Forest Prod J.,1996,46(11/12):27-32
43. Faust T D. Real Time Measurement of Veneer Surface Rough-ness by Image Analysis [J]. Forest Prod.J.1987,37(6):34-40
44. Forrer J B, Funck J W. Dielectric Properties of Defects on Wood Surface [J]. Holzals Rohund Werkstoff.1998,56(1):25-29
45. Gonzalo A, Ruz, Pablo A, Estevez, Claudio A, Perez. A neurofuzzy color image segmentation method for wood surface defect detection [J]. Forest Products Journal,2006,60(4)
46. H.Elbehiery, A.Hefnawy, M.Elewa. Surface Defects Detection for Ceramic Tiles Using Image Processing and Morphological Techniques. Transactions on Engineering Computing and Technology,2005,5:158-162
47. I.J. Jolliffe. Principal Component Analysis [M]. Springer, New York,2002
48. Kenneth R, Castleman. Digital Image Processing[M].Publishing House of Electronic Industry,2005(7):430-432
49. K.Terada,N.Fujiwara, "Extraction of a smoke region using fractal coding", IEEE International Symposium on Communications and Information Technology,2004,ISCIT 2004,Volume 2,26-29 Oct.2004:659-662.
50. L emaster R, Beall C. The Use Optical Profilometer to Measure Surface Roughness in Mediu-m Density Fiberboard [J]. Forest Prod J.,1996,46(11/12):73-78
51. LawrenceK. Saul, SamT. Roweis, Think globally fit locally unsupervised learning of low dimensional manifolds [J]. Journal of Machine Learning Research 2003(4):179-155
52. Mark S. Nixon,Alberto S. Aguado. Feature Extraction and Image Processing [M]. Great Britain: Reed Educational and Professional Publishing Ltd,2006
53. MilanSonka,VaclavHlavac.图像处理、分析与机器视觉[M]2版.北京：人民邮电出版社,2002
54. Pckkarinen A. A Method for the Segmentation of Very High Spatial Resolution Images of Forested Landscapes[J].int.J.Remote Sensing,2002,23(14):2817-2836
55. Proceedings of the 15th International Wood Machining Seminar.2001:473-476
56. SONG J H. Maximum entropy thresholding image segmentation based on genetic algorithm [J].Electronic Engineer,2005,2(9):34-36
57. TAMURA H, MORI S, YAMAWAKI T. Texture features corresponding to visual perception [J]. IEEE Transactions on Systems, Man, and Cybernetics,2003,8(6):460-473
58. WANG X P, CAO L M. Genetic algorithms-theory, application and software implementation [M].Xi'an:Xi'an Jiao tong university press,2002
© 2004-2018 中国地质图书馆版权所有 京ICP备05064691号 京公网安备11010802017129号 地址：北京市海淀区学院路29号 邮编：100083 电话：办公室：(+86 10)66554848；文献借阅、咨询服务、科技查新：66554700 |