【作者】 胡涛；

【导师】 郭宝平；

【作者基本信息】 华中科技大学 ， 光学工程， 2009， 博士

【摘要】 印刷电路板(PCB)是各种电子元器件的载体,在电子通信等许多领域中都有着广泛的应用。随着技术的不断发展以及工艺水平的不断提高,电子产品趋于更轻、更薄、更小,PCB朝着高密度、小元件、细间距、层数更多的方向发展,这使得PCB的质量检验工作越来越具有挑战性。传统的人工检测方法容易漏检、检测速度慢、检测时间长,已经不能满足生产的需要,如何更有效的实现PCB的自动缺陷检测,成为半导体工业领域一个热门问题。自动光学检测技术(AOI)则作为一种新兴的检测技术越来越多的被应用于PCB的缺陷检测。AOI是一种机器视觉技术,集电子学、光电探测、图像处理和计算机技术于一身,是精密测试技术领域内最具有发展潜力的新技术。图像处理软件是AOI最重要的核心组成部分,而软件的核心则主要是检测算法。近年来出现的诸多PCB图像缺陷检测算法大致可分为3类:参考比较法、非参考校验法以及混合算法。其中,参考比较法是目前所普遍采用的,但现有的参考法大都是基于内容的对比,这种方式的检测精度通常只能达到像素级的;此外,由于需要在配准后对图像进行逐点重采样,因而是比较耗时的,当检测图像分辨率较高尺度较大时,所需的处理时间将非常长;再者,其缺陷分析的方式不够直观,某些特定种类的缺陷分析起来比较困难。本论文针对PCB裸板缺陷在线高精度检测的需求,结合由高性能线阵扫描相机获取到的PCB裸板图像所具有的高分辨率以及高对比度等特点,研究并提出一种新的基于轮廓对比的图像缺陷检测算法。基于轮廓对比的图像缺陷检测算法属于参考算法,其所具体包含的内容有图像分割与区域标记、轮廓提取、轮廓亚像素校正、轮廓配准以及缺陷检测与分析。本论文在这几个方面作了逐一分析并提出了相应的解决方案,并在以下几个方面具有创新:(1)提出了一种基于游程连通分析的区域标记算法,只需一次扫描便可实现连通区域的快速标记及目标区域特征的快速提取。(2)提出了一种基于游程连通分析的轮廓提取算法,实现了内外轮廓以及内外轮廓深度包含关系的快速提取。(3)提出了一种基于轮廓点法向角特征的图像配准算法,实现了参考图像与缺陷图像轮廓之间的精确配准。(4)提出了基于轮廓对比的缺陷检测原理,采用轮廓对比的方式实现了各种常见的PCB裸板缺陷识别与检测。基于轮廓对比的检测方式与基于内容对比的检测方式相比有如下优点:首先,由于轮廓数据相对较少,因而可以大大压缩数据存储空间,其次是精度高,通过提取亚像素级的轮廓进行比对分析可将检测精度提高到亚像素级别;第三是检测速度快,这点主要得益于所采用的轮廓对比的方式,无需进行图像重采样,且轮廓数据相对较少,因而计算量大为减少:第四,轮廓对比的方式使得最终的缺陷分析变得更加简单和直观。更多还原

【Abstract】 Printed Circuit Board,a kind of information carrier which integrates varieties of electronic devices,has popular applications in electronic fields nowadays.With the rapid development of manufacture techniques,electronic product tends to lighter,thinner and smaller,and PCB turns to have more layers and higher density,which makes quality detection of PCB become more difficult.Traditional detection methods cannot satisfy the large production for inaccurate,slow and long time detection,then how to realize automated defect detection of PCB becomes a hot topic in semiconductor industry.AOI is a kind of machine vision technology,which combines electronics,photoelectric detection,image processing and computer technology into oneself,is a potential new technology in industrial detection field.Image processing software is the most important component of AOI,and the core of which is inspection algorithm.So far the existing inspection algorithms can be classified into three categories:reference comparison,nonreferential approaches,and hybrid approaches, among which the reference comparison algorithms are the one widely in use now.Most reference comparison algorithms are based on the comparison of content,which have three limitations:firstly,the inspection precision only achieves pixel-level accuracy,secondly,the resampling after image registration is high time-consuming,it will cost very long time to inspect those images with high resolution,thirdly,the way of defect analysis is not direct enough to analyze some special defects.Combining the features of high resolution and high contrast owned by PCB bare board images which are acquired by high performance linear CCD camera,a new inspection algorithm based on contour comparison is proposed to satisfy the real-time and high accuracy defect inspection of PCB bare board.The new inspection algorithm is belong to reference comparison algorithms and consists of image segmentation and region labeling, boundary extraction,sub-pixel boundary computation,contour registration and defect inspection and analysis.The six steps are discussed in sequence and the corresponding solutions have been given,and the innovations lie in the following:(1) A new region labeling algorithm based on run length connectivity analysis is proposed to realize the fast labeling of connected region and the fast extraction of object description just by scanning image data once.(2) A new run based boundary extraction algorithm is proposed to realize the fast and exact extraction of outer and inner boundaries and the deep relationships between them.(3) A new contour normal angle feature based image registration algorithm is proposed to realize exact registration of the contours of the reference image and defect image.(4) The contour comparison theory is proposed to realize the common defects inspection of PCB bare board.Comparing with those content based inspection algorithms,the new contour comparison based inspection algorithm has the following advantages:firstly,less data storage is needed because of the fewer contour data,secondly,higher precision can be achieved by subpixel technology,thirdly,faster inspection speed can be achieved because the contour comparison way does not need the resampling and owns the fewer data size,finally, the contour comparison way make the defect analysis more direct and easy.更多还原