PCB数字图像检测与识别研究
摘要
早期的图像处理主要用于改善图像质量，随着计算机技术、图像采集技术的飞速发展，图像处理也越来越广泛地用于解决众多科学与工程领域中的问题，诸如医学图像分析、指纹鉴别、产品质量检验等。由于计算机图像处理具有许多独特的优点，因而发展非常迅速，成为现代技术研究的热点。
在电子工业中，PCB是各种电子产品的主要部件，PCB上的缺陷可能导致整块印制板甚至整台仪器报废，因此印制板的检验和测试是PCB厂商质量控制不可缺少的环节。
随着生产技术的提高，PCB制作工艺日趋复杂，大量超微器件和超细走线的采用，使PCB的质量检验成为一件非常困难的工作，难以实现自动化。传统的电测法和人工目测法已不能满足生产的需要。运用图像处理和识别技术对PCB图像进行分析，从而找出PCB上存在的缺陷是一种实时、无损、高精度和低成本的优秀方法。目前，市场上已有成品设备销售，通常被称为自动光学检查系统(AOI)，但这些设备多为国外研制，价格昂贵，国内尚在起步阶段；此外，现有的AOI系统仍存在较高的误判率。这说明在PCB图像处理和模式识别方法上仍需进一步的研究。
基于上述几点，本论文以数字图像处理和模式识别为理论基础，对PCB图像的采集、处理、识别的方法和难点进行了深入的探讨，并采用Borland 以C++ Builder6.0编制了一套PCB图像检测与识别软件，对印制板上存在的缺陷种类和数量实现自动检测和识别。
PCB图像检测与识别系统由四部分组成：图像采集、图像预处理、特征提取和图像识别。在图像预处理中，本文分析了常用图像增强和恢复方法对PCB图像的适用性，针对PCB图像的偏暗、对比度不强等特点给出了适合PCB图像特点的处理办法。对图像的二值化，采用双峰法和最大方差阈值法求阈值，取得了较好的效果。在图像识别中，本文分析了常用的模式识别方法，根据树分类法和序贯概率比检定法的思想设计了一种独特的PCB缺陷模式识别方法；并给出了针对各种PCB缺陷模式的特征选择与提取方法。同时，文中还对图像采集系统的设计和采集中的难点提出了解决方案。
在软件研制完成后，对大量PCB图像进了检测，经过对比试验证明能有效地识别印制板上的各种缺陷，收到了较好的效果。论文的设计是成功的，达到了预定的效果，研究成果具有较好的应用价值。
With the development of computer and image acquisition technology, image processing extends its primitive applying in improving quality of image to resolving problems in many fields of science and technology, such as medical image analysis, fingerprint identification, quality inspection of products, and so on. Because of many unique advantages of digital image processing, it develops very rapidly, and now is one of hotspots in modern technology research.
In electronic industry, PCB is the main part of kinds of electronic products, faults of which may led to rejection of the PCB, even more, the whole instrument. So the detecting and testing measure on PCB is necessary for quality control for PCB manufacturer.
Because of the complicated manufacture technique of PCB, for example, a great deal of use of ultro-microcomponent and ultro-slimwire, the quality inspection of PCB becomes very difficult and hardly automatized. Traditional measurement of electronic detection and eye-detection can't satisfy the need in process of manufacture. The way that faults on PCB are found by means of image processing and recognition technique is an excellent method, real-time, nondestructive, high-accuracy and low cost. Now some instruments have been on sale, which is called the Automatic Optical Inspection System(AOI). However, AOI is mainly made by foreign producer with expensive price, while its research is still underway in China; Meanwhile, AOI exists some disadvantages such as high misjudgement rate. All show that the method of PCB image processing and pattern recognition needs more studying.
As the result, based on the principles of digital image processing and pattern recognition, this paper is focused on the methods and difficulties of acquisition, processing, recognition of PCB. With the use of Borland C++ Builder 6.0, a set of PCB image detection and recognition software system was developed, which can realize automatic detection and recognition for types and the amount of faults on PCB.
The whole PCB image detection & recognition system consists of 4 parts: image acquisition; image pre-processing; feature extraction; image recognition. In terms of dim, low contrast of PCB image at the stage of image pre-processing, by analyzing the applicability of common methods of image enhancing and resuming for PCB image, a certain resolving methods have been shown: Obtaining the thresholding value by means of 2-mode method and maximum variance thresholding method, the result of image binarizaion was satisfied. At the stage of image recognition, a unique model of PCB fault recognition was built on methods of tree-classification and sequenntial probability ratio test, and a kind of method of
m
feature selection and extraction was introduced. In the same time, resolving methods for difficulties in design and acquisition of PCB image acquisition system were given out.
Using this software system, amounts of PCB image have been tested. By comparing tests, the results show that it can recognize kinds of faults on PCB effectively, which proves the design of this paper and thus it has practical value.
在电子工业中，PCB是各种电子产品的主要部件，PCB上的缺陷可能导致整块印制板甚至整台仪器报废，因此印制板的检验和测试是PCB厂商质量控制不可缺少的环节。
随着生产技术的提高，PCB制作工艺日趋复杂，大量超微器件和超细走线的采用，使PCB的质量检验成为一件非常困难的工作，难以实现自动化。传统的电测法和人工目测法已不能满足生产的需要。运用图像处理和识别技术对PCB图像进行分析，从而找出PCB上存在的缺陷是一种实时、无损、高精度和低成本的优秀方法。目前，市场上已有成品设备销售，通常被称为自动光学检查系统(AOI)，但这些设备多为国外研制，价格昂贵，国内尚在起步阶段；此外，现有的AOI系统仍存在较高的误判率。这说明在PCB图像处理和模式识别方法上仍需进一步的研究。
基于上述几点，本论文以数字图像处理和模式识别为理论基础，对PCB图像的采集、处理、识别的方法和难点进行了深入的探讨，并采用Borland 以C++ Builder6.0编制了一套PCB图像检测与识别软件，对印制板上存在的缺陷种类和数量实现自动检测和识别。
PCB图像检测与识别系统由四部分组成：图像采集、图像预处理、特征提取和图像识别。在图像预处理中，本文分析了常用图像增强和恢复方法对PCB图像的适用性，针对PCB图像的偏暗、对比度不强等特点给出了适合PCB图像特点的处理办法。对图像的二值化，采用双峰法和最大方差阈值法求阈值，取得了较好的效果。在图像识别中，本文分析了常用的模式识别方法，根据树分类法和序贯概率比检定法的思想设计了一种独特的PCB缺陷模式识别方法；并给出了针对各种PCB缺陷模式的特征选择与提取方法。同时，文中还对图像采集系统的设计和采集中的难点提出了解决方案。
在软件研制完成后，对大量PCB图像进了检测，经过对比试验证明能有效地识别印制板上的各种缺陷，收到了较好的效果。论文的设计是成功的，达到了预定的效果，研究成果具有较好的应用价值。
With the development of computer and image acquisition technology, image processing extends its primitive applying in improving quality of image to resolving problems in many fields of science and technology, such as medical image analysis, fingerprint identification, quality inspection of products, and so on. Because of many unique advantages of digital image processing, it develops very rapidly, and now is one of hotspots in modern technology research.
In electronic industry, PCB is the main part of kinds of electronic products, faults of which may led to rejection of the PCB, even more, the whole instrument. So the detecting and testing measure on PCB is necessary for quality control for PCB manufacturer.
Because of the complicated manufacture technique of PCB, for example, a great deal of use of ultro-microcomponent and ultro-slimwire, the quality inspection of PCB becomes very difficult and hardly automatized. Traditional measurement of electronic detection and eye-detection can't satisfy the need in process of manufacture. The way that faults on PCB are found by means of image processing and recognition technique is an excellent method, real-time, nondestructive, high-accuracy and low cost. Now some instruments have been on sale, which is called the Automatic Optical Inspection System(AOI). However, AOI is mainly made by foreign producer with expensive price, while its research is still underway in China; Meanwhile, AOI exists some disadvantages such as high misjudgement rate. All show that the method of PCB image processing and pattern recognition needs more studying.
As the result, based on the principles of digital image processing and pattern recognition, this paper is focused on the methods and difficulties of acquisition, processing, recognition of PCB. With the use of Borland C++ Builder 6.0, a set of PCB image detection and recognition software system was developed, which can realize automatic detection and recognition for types and the amount of faults on PCB.
The whole PCB image detection & recognition system consists of 4 parts: image acquisition; image pre-processing; feature extraction; image recognition. In terms of dim, low contrast of PCB image at the stage of image pre-processing, by analyzing the applicability of common methods of image enhancing and resuming for PCB image, a certain resolving methods have been shown: Obtaining the thresholding value by means of 2-mode method and maximum variance thresholding method, the result of image binarizaion was satisfied. At the stage of image recognition, a unique model of PCB fault recognition was built on methods of tree-classification and sequenntial probability ratio test, and a kind of method of
m
feature selection and extraction was introduced. In the same time, resolving methods for difficulties in design and acquisition of PCB image acquisition system were given out.
Using this software system, amounts of PCB image have been tested. By comparing tests, the results show that it can recognize kinds of faults on PCB effectively, which proves the design of this paper and thus it has practical value.
引文
1．王积分，张新荣．计算机图象识别．中国铁道出版社，1988
2．赵荣椿等．数字图象处理导论．西北工业大学出版社，1995
3．荆仁杰等．计算机图像处理．浙江大学出版社，1990
4．R.C.冈萨雷斯，P.温茨．数字图像处理．李叔梁等译．科学出版社，1983
5．沈清，汤霖．模式识别导论．国防科技大学出版社，1991
6．傅京孙．模式识别及其应用．科学出版社，1983
7．李月景．图像识别技术及其应用，机械工业出版社，1985
8．黄智．图像处理和识别实用程序库，天津科学技术出版社，1989
9．卜家岐，林贻侠．C程序设计教程．中国科学技术出版社，1995
10．郑莉，董渊．C++语言程序设计(第2版)．清华大学出版社，2001
11．余明兴等．C++ Builder 5程序设计．中国铁道出版社，2001
12．刘滨．C++ Builder 5高级编程实例精解，国防工业出版社，2001
13．清宏计算机工作室．C++ Builder 编程技巧(多媒体与系统篇)．机械工业出版社，2001
14．蔡文贵等．CCD技术及应用．电子工业出版社，1992
15．Adobe公司北京代表处．Adobe Photoshop 7.0标准培训教材．人民邮电出版社，2002
16．崔屹．数字图象处理技术与应用．电子工业出版社，1997
17．夏良正．数字图像处理．东南大学出版社，1999
18．容观澳．计算机图像处理．清华大学出版社，2000
19．刘华等．Borland C++ Builder程序设计．清华大学出版社，2001
20．Bill Erickson & Frank Romano. Professional Digital Photography.承健等译．电子工业出版社，1999
21．王格芳等．基于红外热成像的集成电路检测与诊断．激光与红外．1999，29(5)
22．王秋成等．印刷电路板电脑检测仪的设计．机电工程．1999，16(1)
23．卢朝阳等．用图像识别的方法检测集成电路的键合点．自动化学报．1999，25(4)
24．戴明远．印刷电路板短路处的寻找方法．电测与仪表．2000，37(1)
25．周铭等．基于视频图形图像迭加的印刷线路板定位孔钻机．1999，8
26．刘兴占，巩马理，李伦．CCD扫描法检测模具表面缺陷的研究．光学技术．2000，26(4)
27．刘恒坤，佘龙华等，CCD线阵在姿态检测中的应用．传感器技术．2000，19(5)
28．李庆中等．基于计算机视觉的苹果自动分级系统硬件开发．农业机械学报．2000，31(2)
29．陈鹰，叶勤，钟志勇．基于小波变幻的雷达与光学影像匹配算法研究．测绘学报．2000，29(3)
30．李全臣，夏亮．成像光谱的信息采集与数据处理．光学技术．1999，(3)
31．王金刚，李伟，刘颖．全帧读出型面阵CCD光电传感器在图像采集中的应用．电子技术应用．2000，7
32．冯勇，韩凤岭等．TDI CCD成像系统的研究．光学技术．2000，26(2)
33．王学军．激光和CCD快速自动测量车身曲面的研究．应用激光．1998，18(6)
34．何东建．果实缺陷面积的计算机视觉测定研究．农业工程学报．1997，13(4)
35．朱小平，肖明清．基于增强型并行口EPP协议的PCB故障诊断系统．测控技术．2002，21(7)
36．沈奕，陆霖琳等．LCD电极检测中定位标记的图像识别．电子技术应用．2001，27(9)
37. Sophocles J. Orfanidis. Introduction to Signal Processing. Prentice Hall Inc.,1996
38. A.K. Jain. Fundamentals of Digital Image Processing. Prentice Hall Inc, 1989
39. J. Kittler, J. Illingworth. Threshold Selection Based on a Simple Image Stastic. Computer Vision, Graphics and Image Processing. 1985, 30
40. Laveen N. Kanal. Progress in Pattern Recognition. North-Holland Publishing comp.,1981
41. K. R. Castleman. Digital Image Processing. 2nd Ed. Prentice Hall, 1996
42. G. E. Revese. Inroduction to Formal Languages. McGraw-Hill comp., 1983
43. Photometrics Ltd..Signal Processing and Noise. Series 200CCD Cameras Manual, 1990
44. N. J. Nilson. Principles of Artifical Intelligence. Tioga
Publishing comp.,1980
45. T. S. Newman, A. K. Jain. A Survey of Automated Visual inspection. Computer Vision and Image Understanding. 1995,61(2):213-262.
46. J. N. Kapur, P. K. Sahoo, etc. A New Method for Gray-level Picture Thresholding Using the Entropy of the Histogram. Computer Vision. Graphics and Image Processing. 1985,29(3):273-285.
47. Wang Zhenqing, Shen Qing, Liu Shaochu. The Intellectualized Direction of Pattern Recognition. ICCPOL 90, Apr., 1990
48. J. L. McClelland, D. E. Rumelhart. Explorations in Parallel Distributed Processing. Handbook of Models. Vol. 12, 1989
49. R. P. Lippman. Pattern Classification Using Neural Networks. IEEE Communications Magazine, Nov. 1989
50. L. Vincent. Morphological Transformations of Binary Images with Arbitrary Structuring Elements. Singal Processing. 1991, 22(1)
51. J. Canny. A Computational Approach to Edge Detection. IEEE Transactions on Pattern Analysis and Machine Intelligence. 1986, PAMI-8(6)
52. P. W. Verbeek, L. J. Van Vliet. On the Location Error of Curved Edges in Low-Pass Filtered 2D and 3D Images. IEEE Transactions on Pattern Analysis and Machine Intelligence. 1994,16(7)
53. L. J. Van Vliet, I.T. Young, A. L. D. Beckers. A Non-Linear Laplace Operator as Edge Detector in Noisy Images. Computer Vision, Graphics and Image Processing. 1989, 45
54．宁静．数字图像识别在测控系统中的应用．西南交通大学硕士论文．2002
2．赵荣椿等．数字图象处理导论．西北工业大学出版社，1995
3．荆仁杰等．计算机图像处理．浙江大学出版社，1990
4．R.C.冈萨雷斯，P.温茨．数字图像处理．李叔梁等译．科学出版社，1983
5．沈清，汤霖．模式识别导论．国防科技大学出版社，1991
6．傅京孙．模式识别及其应用．科学出版社，1983
7．李月景．图像识别技术及其应用，机械工业出版社，1985
8．黄智．图像处理和识别实用程序库，天津科学技术出版社，1989
9．卜家岐，林贻侠．C程序设计教程．中国科学技术出版社，1995
10．郑莉，董渊．C++语言程序设计(第2版)．清华大学出版社，2001
11．余明兴等．C++ Builder 5程序设计．中国铁道出版社，2001
12．刘滨．C++ Builder 5高级编程实例精解，国防工业出版社，2001
13．清宏计算机工作室．C++ Builder 编程技巧(多媒体与系统篇)．机械工业出版社，2001
14．蔡文贵等．CCD技术及应用．电子工业出版社，1992
15．Adobe公司北京代表处．Adobe Photoshop 7.0标准培训教材．人民邮电出版社，2002
16．崔屹．数字图象处理技术与应用．电子工业出版社，1997
17．夏良正．数字图像处理．东南大学出版社，1999
18．容观澳．计算机图像处理．清华大学出版社，2000
19．刘华等．Borland C++ Builder程序设计．清华大学出版社，2001
20．Bill Erickson & Frank Romano. Professional Digital Photography.承健等译．电子工业出版社，1999
21．王格芳等．基于红外热成像的集成电路检测与诊断．激光与红外．1999，29(5)
22．王秋成等．印刷电路板电脑检测仪的设计．机电工程．1999，16(1)
23．卢朝阳等．用图像识别的方法检测集成电路的键合点．自动化学报．1999，25(4)
24．戴明远．印刷电路板短路处的寻找方法．电测与仪表．2000，37(1)
25．周铭等．基于视频图形图像迭加的印刷线路板定位孔钻机．1999，8
26．刘兴占，巩马理，李伦．CCD扫描法检测模具表面缺陷的研究．光学技术．2000，26(4)
27．刘恒坤，佘龙华等，CCD线阵在姿态检测中的应用．传感器技术．2000，19(5)
28．李庆中等．基于计算机视觉的苹果自动分级系统硬件开发．农业机械学报．2000，31(2)
29．陈鹰，叶勤，钟志勇．基于小波变幻的雷达与光学影像匹配算法研究．测绘学报．2000，29(3)
30．李全臣，夏亮．成像光谱的信息采集与数据处理．光学技术．1999，(3)
31．王金刚，李伟，刘颖．全帧读出型面阵CCD光电传感器在图像采集中的应用．电子技术应用．2000，7
32．冯勇，韩凤岭等．TDI CCD成像系统的研究．光学技术．2000，26(2)
33．王学军．激光和CCD快速自动测量车身曲面的研究．应用激光．1998，18(6)
34．何东建．果实缺陷面积的计算机视觉测定研究．农业工程学报．1997，13(4)
35．朱小平，肖明清．基于增强型并行口EPP协议的PCB故障诊断系统．测控技术．2002，21(7)
36．沈奕，陆霖琳等．LCD电极检测中定位标记的图像识别．电子技术应用．2001，27(9)
37. Sophocles J. Orfanidis. Introduction to Signal Processing. Prentice Hall Inc.,1996
38. A.K. Jain. Fundamentals of Digital Image Processing. Prentice Hall Inc, 1989
39. J. Kittler, J. Illingworth. Threshold Selection Based on a Simple Image Stastic. Computer Vision, Graphics and Image Processing. 1985, 30
40. Laveen N. Kanal. Progress in Pattern Recognition. North-Holland Publishing comp.,1981
41. K. R. Castleman. Digital Image Processing. 2nd Ed. Prentice Hall, 1996
42. G. E. Revese. Inroduction to Formal Languages. McGraw-Hill comp., 1983
43. Photometrics Ltd..Signal Processing and Noise. Series 200CCD Cameras Manual, 1990
44. N. J. Nilson. Principles of Artifical Intelligence. Tioga
Publishing comp.,1980
45. T. S. Newman, A. K. Jain. A Survey of Automated Visual inspection. Computer Vision and Image Understanding. 1995,61(2):213-262.
46. J. N. Kapur, P. K. Sahoo, etc. A New Method for Gray-level Picture Thresholding Using the Entropy of the Histogram. Computer Vision. Graphics and Image Processing. 1985,29(3):273-285.
47. Wang Zhenqing, Shen Qing, Liu Shaochu. The Intellectualized Direction of Pattern Recognition. ICCPOL 90, Apr., 1990
48. J. L. McClelland, D. E. Rumelhart. Explorations in Parallel Distributed Processing. Handbook of Models. Vol. 12, 1989
49. R. P. Lippman. Pattern Classification Using Neural Networks. IEEE Communications Magazine, Nov. 1989
50. L. Vincent. Morphological Transformations of Binary Images with Arbitrary Structuring Elements. Singal Processing. 1991, 22(1)
51. J. Canny. A Computational Approach to Edge Detection. IEEE Transactions on Pattern Analysis and Machine Intelligence. 1986, PAMI-8(6)
52. P. W. Verbeek, L. J. Van Vliet. On the Location Error of Curved Edges in Low-Pass Filtered 2D and 3D Images. IEEE Transactions on Pattern Analysis and Machine Intelligence. 1994,16(7)
53. L. J. Van Vliet, I.T. Young, A. L. D. Beckers. A Non-Linear Laplace Operator as Edge Detector in Noisy Images. Computer Vision, Graphics and Image Processing. 1989, 45
54．宁静．数字图像识别在测控系统中的应用．西南交通大学硕士论文．2002