LED晶粒分拣技术的机器视觉研究
|
摘要
LED晶粒的分拣是LED生产中的一个重要环节,由于分拣设备长久以来一直被国外市场垄断,所以研发具有自主知识产权的LED分拣设备势在必行。而作为分拣设备中,机器视觉技术又是至关重要的,它设计的好坏直接影响着整个分拣的系统。本文主要研究机器视觉技术在LED分拣设备中的应用。
论文以高速CCD相机、镜头、图像采集卡及工控机为核心组件,构建了基于机器视觉的硬件平台,用来对图像进行采集与处理。在LED分拣设备中,晶粒的位置信息是至关重要的,它为整个的分拣工作提供了基本参数信息。为了得到LED晶粒的位置参数,论文中采用了基于区域灰度值的模板匹配算法。深入分析了模板匹配算法的功能及预处理的过程。在相应的匹配工作基础上要对图像进行区域的分割,进而改变图像的表示方法,并对图像像素进行再组织,形成更高级的表示单元。区域分割是一个集合划分的过程,它只是把相似的像素划分到不同的区域中,分割得到的区域仍然有很多冗余信息为了消除这些冗余信息,在此提出了感兴趣区域来描述图像。它可以增强图像对比度,能显著的描述物体的轮廓和形状,并在一定程度上消除了图像的冗余信息,突出图像的主要内容。CCD在拍摄过程中,由于光学镜头存在有一定的光学畸变,将会引入一定的测量误差,从而影响晶粒的识别和定位。为了解决这一问题,本文提出采用预处理前对图像进行校正的措施来消除畸变。识别完成后会生成LED位置信息的参数表,该表为运动控制提供了运动数据。
本课题以VC++6.0作为软件开发平台,使用Sapera Processing图像处理软件来实现开发LED分拣设备的机器视觉的软件控制系统。利用机器视觉的处理算法来实现图像的识别与定位。经过测试结果表明,论文中所构建的硬件平台及机器视觉处理算法合理,且处理精度达到生产精度的要求,具有很好的实用性。
The Sorting of LED is one of the most components of the LED production. In the past, sorting equipment has been monopolized by foreign companies; therefore, it's extremely urgent and necessary to develop LED sorting equipment with independent intellectual property rights. In sorting equipment, machine vision technology is of paramount important, whose design have a significant impact on the whole sorting system. This paper mainly focused on the application of machine vision technology into LED sorting equipment.
The machine vision hardware platform is established to gather and process image base on high-speed CCD camera, lens, image acquisition card and industrial control computer in this paper. In the LED sorting equipment, the dies location information is essential and works for the entire basic parameter information. The template matching algorithm based on gray value of region was developed to determine the LED dies' level and location parameters. This study further analyzed the function and pretreatment process of the template matching algorithm. Before the corresponding matching, the image was regionally segmented. The main objective of segmentation was to divide the image into small components, change the way of image presentation and to re-organize image pixels to form a more advanced unit, which was also useful to facilitate in-depth analysis. Regional segmentation was only a partitioning process; and it only put similar pixels into different regions; As a result, there were still much redundant information after segmentation. Therefore, we would like to describe the image with regions of interest (ROI). ROI refers to the image region which can arouse the interest of the most users' and is able to well present main image content; it can exactly show the profile and shape of the objects; it can eliminate redundant image information in a certain extent and highlight the core content of the image. Because optical distortion that caused by lens affect the identification and positioning of dies when CCD is working. To solve this problem, the measure of pre-process was applied before the image correction. After identification, it will generate LED location parameter table, which provides for the motion control the movement of data.
Base on VC + +6.0 as a software development platform, Sapera Processing image processing software was applied to develop the software control system of machine vision for LED sorting equipment. Machine vision processing algorithms was developed to achieve the image recognition and location. The test results show that the hardware platform and visual processing algorithms in this study are reasonable, and the processing accuracy can meet the requirements of production. The system has good practicability and utility.
论文以高速CCD相机、镜头、图像采集卡及工控机为核心组件,构建了基于机器视觉的硬件平台,用来对图像进行采集与处理。在LED分拣设备中,晶粒的位置信息是至关重要的,它为整个的分拣工作提供了基本参数信息。为了得到LED晶粒的位置参数,论文中采用了基于区域灰度值的模板匹配算法。深入分析了模板匹配算法的功能及预处理的过程。在相应的匹配工作基础上要对图像进行区域的分割,进而改变图像的表示方法,并对图像像素进行再组织,形成更高级的表示单元。区域分割是一个集合划分的过程,它只是把相似的像素划分到不同的区域中,分割得到的区域仍然有很多冗余信息为了消除这些冗余信息,在此提出了感兴趣区域来描述图像。它可以增强图像对比度,能显著的描述物体的轮廓和形状,并在一定程度上消除了图像的冗余信息,突出图像的主要内容。CCD在拍摄过程中,由于光学镜头存在有一定的光学畸变,将会引入一定的测量误差,从而影响晶粒的识别和定位。为了解决这一问题,本文提出采用预处理前对图像进行校正的措施来消除畸变。识别完成后会生成LED位置信息的参数表,该表为运动控制提供了运动数据。
本课题以VC++6.0作为软件开发平台,使用Sapera Processing图像处理软件来实现开发LED分拣设备的机器视觉的软件控制系统。利用机器视觉的处理算法来实现图像的识别与定位。经过测试结果表明,论文中所构建的硬件平台及机器视觉处理算法合理,且处理精度达到生产精度的要求,具有很好的实用性。
The Sorting of LED is one of the most components of the LED production. In the past, sorting equipment has been monopolized by foreign companies; therefore, it's extremely urgent and necessary to develop LED sorting equipment with independent intellectual property rights. In sorting equipment, machine vision technology is of paramount important, whose design have a significant impact on the whole sorting system. This paper mainly focused on the application of machine vision technology into LED sorting equipment.
The machine vision hardware platform is established to gather and process image base on high-speed CCD camera, lens, image acquisition card and industrial control computer in this paper. In the LED sorting equipment, the dies location information is essential and works for the entire basic parameter information. The template matching algorithm based on gray value of region was developed to determine the LED dies' level and location parameters. This study further analyzed the function and pretreatment process of the template matching algorithm. Before the corresponding matching, the image was regionally segmented. The main objective of segmentation was to divide the image into small components, change the way of image presentation and to re-organize image pixels to form a more advanced unit, which was also useful to facilitate in-depth analysis. Regional segmentation was only a partitioning process; and it only put similar pixels into different regions; As a result, there were still much redundant information after segmentation. Therefore, we would like to describe the image with regions of interest (ROI). ROI refers to the image region which can arouse the interest of the most users' and is able to well present main image content; it can exactly show the profile and shape of the objects; it can eliminate redundant image information in a certain extent and highlight the core content of the image. Because optical distortion that caused by lens affect the identification and positioning of dies when CCD is working. To solve this problem, the measure of pre-process was applied before the image correction. After identification, it will generate LED location parameter table, which provides for the motion control the movement of data.
Base on VC + +6.0 as a software development platform, Sapera Processing image processing software was applied to develop the software control system of machine vision for LED sorting equipment. Machine vision processing algorithms was developed to achieve the image recognition and location. The test results show that the hardware platform and visual processing algorithms in this study are reasonable, and the processing accuracy can meet the requirements of production. The system has good practicability and utility.
引文
[1]斯白露.基于感兴趣区域的图像检索方法[D].北京：中国科学院研究生院,2002.5：12
[2]尚艳丽.LED芯片检测和分选的图像匹配算法研究[D].湖北：华中科技大学.2007
[3]王艳娟.基于感兴趣区域的图像检索技术研究[D].阜新：辽宁工程技术大学,2004.11
[4]蔡睿.基于DSP的数字图像畸变校正技术研究[D].武汉：华中科技大学,2006.7
[5]让星.基于机器视觉的密封件图像测量技术研究[D].西安：西安工业大学,2009.5：21-24
[6]Carsten Steger, Markus Ulrich, Christian Wiedemann.Machine Vision Algorithms杨少容,吴迪靖,段德山(译).机器视觉算法与应用[M].北京：清华大学出版社,2008.1：302-308
[7]吴学明.图像分割的算法研究[D].四川：成都理工大学.2006
[8]何东健.数字图像处理.西安电子科技大学出版社,2003.7：90-111
[9]乔玲玲.图像分割算法研究及实现[D].湖北：武汉理工大学2006
[10]管慧娟.基于区域的图像分割方法[D].大连：大连理工大学,2005
[11]曹占伦,吴小洪,姜永军等.改进匹配算法在LED粘片设备图像识别中的应用[J／OL].半导体技术第33卷第3期2007-11-21
[12]朱欣昱,郗安民,伊文君.新型二极管芯片分拣机及分拣方法[J／OL].研究机械设计与制造,2007年第06期.
[13]李兴南,范德斌,陈剑军等.摄像机标定算法研究[J／OL].九江学院学报,2009年03期
[14]楼兵军.基于区域的图像匹配算法的关键技术研究[D].西安：西安电子科技大学,2006.2
[15]刘传才.图像理解与计算机视觉[M].厦门：厦门大学出版社,2002：114-127
[16]高和蓓.基于感兴趣区域的图像检索[D].扬州：扬州大学,2006.5：33-35
[17]杜维涛.基于感兴趣区域的遥感图像压缩技术[D].长沙：国防科学技术大学研究生院,2007.11：25-38
[18]李超.静态图像感兴趣区域提取关键技术研究[D].天津：天津大学,2007.6：5-9
[19]于泓,陈辉,赵辉.基于交叉垂直线的摄像机标定新算法[J／OL].计算机应用,2006年01期
[20]姜文瀚.模式识别中的样本选择研究及其应用[D].南京：南京理工大学博士学位论文,2007.9
[21]杜召斌.基于单目CCD摄像机的大空间环境下坐标定位[D].成都：西华大学硕士学位论文,2006.11：31-35
[22]杨淑莹.VC++图像处理程序设计[M].北京：清华大学出版社；北京交通大学出版社,2005
[23]马颂德,张正友.计算机视觉[M].北京：科学出版社,2003：11-34.
[24]罗丁,吴小洪,姜石军等.多目标快速搜索算法——芯片图像识别中应用[J／OL].微计算机信息(测控自动化)2006年第22卷第6-1期
[25]孙鑫.VC++深入详解[M].电子工业出版社,2006.
[26]张志佳,张瑜,史泽林等.基于区域模板相关的图像匹配方法研究[J／OL].红外与激光工程,第33卷第6期
[27]阮秋琦.数字图像处理学[M].北京：电子工业出版社,2001：180-480.
[28]齐舒创作室编.Visual C++6.0用户界面制作技术与应用实例[M].中国水利水电出版社,1999
[29]谢涛.基于区域分割的立体匹配及其相关研究[D].大连：大连理工大学,2005.12
[30]郑南宁.计算机视觉与模式识别[M].国防工业出版社,1998
[31]贾云德.机器视觉[M].北京：科学出版社,2000
[32]杨光正,吴岷,张晓莉.模式识别[M].合肥：中国科学技术大学出版社,2001.
[33]孙即祥,王晓华,钟山等.模式识别中的特征提取与计算机视觉不变量[M].北京：国防工业出版社,2001
[34]张新峰,沈兰荪.模式识别及其在图像处理中的应用[J／OL].测控技术2004年第23卷第五期
[35]Zobel M,Gebhard A,Paulus D,et al. Robust facial feature localization by coupled features,Proceedings[J/OL]. Fourth IEEE International Con2ferenceon Automatic Faceand Gesture Recognition,2000-03:2-7.
[36]Kohir V V, Desai U B. Face recognition [M].IEEE International Symposium on Circuit and System,2000-05:305-308.
[37]Sanderson C,Palimal K K.Fast feature extraction method for robust face verification [J/OL].Electronics Letters,2002,38(25).
[38]Ivind Due Trier, Anil K. Jain, Torfinn Taxt. Feature extraction methods for character recognition-a survey[J/OL].Pattern Recognation,1996,29 (4):641-662.
[39]Tang Yuan Y, Tao Yu,Lam Ernest C M. New method for feature extraction based on fractal behavior [J/OL]. Pattern Recognition,2002,55 (5):1071-1081.
[40]M. T. Ahmed,E. Hernayed and A. Farag, Camera calibration with lens distortion using weakly-localized image features[M]. Proc. Of SPIE,365:119-127,1999
[41]Rainer Lienhart and Jochen Maydt. An Extended Set of Hear-like Features for Rapid Object Detection[M]. IEEE ICIP 2002, Vol.1, pp.900-903, Sep.2002.
[42]Paul Viola and Michael J. Jones. Rapid Object Detection using a Boosted Cascade of Simple Features[J/OL]. IEEE CVPR,2001.
[43]G. Bradski and J. Davis. Motion Segmentation and Pose Recognition with Motion History Gradients[M]. IEEE WACV'00,2000.
[44]Chen Li, Yap Kim2Hui. A fuzzy K2nearest2neighbor algorithm to blind image. Deconvolution [J/OL].IEEE International Conference on Systems, Man and Cybernetics,2003-03:2049-2054.
[45]Corona2Nakamura M A,Ruelas R. Improving classification with auto2 mated selection of a combined classifier [M]. Proceedings of the 5th Biannual,2002,13:443-448.
[2]尚艳丽.LED芯片检测和分选的图像匹配算法研究[D].湖北：华中科技大学.2007
[3]王艳娟.基于感兴趣区域的图像检索技术研究[D].阜新：辽宁工程技术大学,2004.11
[4]蔡睿.基于DSP的数字图像畸变校正技术研究[D].武汉：华中科技大学,2006.7
[5]让星.基于机器视觉的密封件图像测量技术研究[D].西安：西安工业大学,2009.5：21-24
[6]Carsten Steger, Markus Ulrich, Christian Wiedemann.Machine Vision Algorithms杨少容,吴迪靖,段德山(译).机器视觉算法与应用[M].北京：清华大学出版社,2008.1：302-308
[7]吴学明.图像分割的算法研究[D].四川：成都理工大学.2006
[8]何东健.数字图像处理.西安电子科技大学出版社,2003.7：90-111
[9]乔玲玲.图像分割算法研究及实现[D].湖北：武汉理工大学2006
[10]管慧娟.基于区域的图像分割方法[D].大连：大连理工大学,2005
[11]曹占伦,吴小洪,姜永军等.改进匹配算法在LED粘片设备图像识别中的应用[J／OL].半导体技术第33卷第3期2007-11-21
[12]朱欣昱,郗安民,伊文君.新型二极管芯片分拣机及分拣方法[J／OL].研究机械设计与制造,2007年第06期.
[13]李兴南,范德斌,陈剑军等.摄像机标定算法研究[J／OL].九江学院学报,2009年03期
[14]楼兵军.基于区域的图像匹配算法的关键技术研究[D].西安：西安电子科技大学,2006.2
[15]刘传才.图像理解与计算机视觉[M].厦门：厦门大学出版社,2002：114-127
[16]高和蓓.基于感兴趣区域的图像检索[D].扬州：扬州大学,2006.5：33-35
[17]杜维涛.基于感兴趣区域的遥感图像压缩技术[D].长沙：国防科学技术大学研究生院,2007.11：25-38
[18]李超.静态图像感兴趣区域提取关键技术研究[D].天津：天津大学,2007.6：5-9
[19]于泓,陈辉,赵辉.基于交叉垂直线的摄像机标定新算法[J／OL].计算机应用,2006年01期
[20]姜文瀚.模式识别中的样本选择研究及其应用[D].南京：南京理工大学博士学位论文,2007.9
[21]杜召斌.基于单目CCD摄像机的大空间环境下坐标定位[D].成都：西华大学硕士学位论文,2006.11：31-35
[22]杨淑莹.VC++图像处理程序设计[M].北京：清华大学出版社；北京交通大学出版社,2005
[23]马颂德,张正友.计算机视觉[M].北京：科学出版社,2003：11-34.
[24]罗丁,吴小洪,姜石军等.多目标快速搜索算法——芯片图像识别中应用[J／OL].微计算机信息(测控自动化)2006年第22卷第6-1期
[25]孙鑫.VC++深入详解[M].电子工业出版社,2006.
[26]张志佳,张瑜,史泽林等.基于区域模板相关的图像匹配方法研究[J／OL].红外与激光工程,第33卷第6期
[27]阮秋琦.数字图像处理学[M].北京：电子工业出版社,2001：180-480.
[28]齐舒创作室编.Visual C++6.0用户界面制作技术与应用实例[M].中国水利水电出版社,1999
[29]谢涛.基于区域分割的立体匹配及其相关研究[D].大连：大连理工大学,2005.12
[30]郑南宁.计算机视觉与模式识别[M].国防工业出版社,1998
[31]贾云德.机器视觉[M].北京：科学出版社,2000
[32]杨光正,吴岷,张晓莉.模式识别[M].合肥：中国科学技术大学出版社,2001.
[33]孙即祥,王晓华,钟山等.模式识别中的特征提取与计算机视觉不变量[M].北京：国防工业出版社,2001
[34]张新峰,沈兰荪.模式识别及其在图像处理中的应用[J／OL].测控技术2004年第23卷第五期
[35]Zobel M,Gebhard A,Paulus D,et al. Robust facial feature localization by coupled features,Proceedings[J/OL]. Fourth IEEE International Con2ferenceon Automatic Faceand Gesture Recognition,2000-03:2-7.
[36]Kohir V V, Desai U B. Face recognition [M].IEEE International Symposium on Circuit and System,2000-05:305-308.
[37]Sanderson C,Palimal K K.Fast feature extraction method for robust face verification [J/OL].Electronics Letters,2002,38(25).
[38]Ivind Due Trier, Anil K. Jain, Torfinn Taxt. Feature extraction methods for character recognition-a survey[J/OL].Pattern Recognation,1996,29 (4):641-662.
[39]Tang Yuan Y, Tao Yu,Lam Ernest C M. New method for feature extraction based on fractal behavior [J/OL]. Pattern Recognition,2002,55 (5):1071-1081.
[40]M. T. Ahmed,E. Hernayed and A. Farag, Camera calibration with lens distortion using weakly-localized image features[M]. Proc. Of SPIE,365:119-127,1999
[41]Rainer Lienhart and Jochen Maydt. An Extended Set of Hear-like Features for Rapid Object Detection[M]. IEEE ICIP 2002, Vol.1, pp.900-903, Sep.2002.
[42]Paul Viola and Michael J. Jones. Rapid Object Detection using a Boosted Cascade of Simple Features[J/OL]. IEEE CVPR,2001.
[43]G. Bradski and J. Davis. Motion Segmentation and Pose Recognition with Motion History Gradients[M]. IEEE WACV'00,2000.
[44]Chen Li, Yap Kim2Hui. A fuzzy K2nearest2neighbor algorithm to blind image. Deconvolution [J/OL].IEEE International Conference on Systems, Man and Cybernetics,2003-03:2049-2054.
[45]Corona2Nakamura M A,Ruelas R. Improving classification with auto2 mated selection of a combined classifier [M]. Proceedings of the 5th Biannual,2002,13:443-448.