【作者】 陈华；

【导师】 白金平；

【作者基本信息】 电子科技大学 ， 系统工程， 2020， 硕士

【摘要】 随着电子产品朝着轻、薄、小的方向发展,柔性印刷电路板(Flexible Printed Circuit,简称FPC)凭借其轻、薄、可挠的优势被运用地越来越广泛。我国是全球最主要的FPC生产基地。FPC产量和本身配线复杂度的提高,使得质检成为FPC生产过程中重要的一环,人工检测FPC缺陷已经远远不能满足企业的要求。本文以某FPC生产厂商提供的一款FPC作为研究对象,该款FPC刚性结构与挠性结构交错,走线相对复杂,具有一定代表性。在研究国内外线路板检测相关文献的基础上,提炼并研究了 FPC检测涉及的两大关键问题,FPC图像分割和FPC图像配准,并在此基础上设计了 FPC表面缺陷检测算法。本文将检测目标分为金属区域和线路区域,根据金属区域亮度高,线路区域分布广的特点分别设计了图像分割算法,针对线路区域的局部阂值分割,通过实验得出两倍最大线路宽度为最佳采样矩形边长的结论。接着本文研究了 FPC的图像配准,为了减弱环境的干扰和减少数据计算量,设计了高效的FPC目标定位算法。在完成定位的基础上,对FPC图像配准进行了研究,设计了基于定位孔圆心特征的图像配准方法,发现图像不能实现全局配准,初步进行了基于局部特征的局部配准方法的研究,得出了 FPC图像变换是非线性的结论,为检测算法的设计奠定了基础。最后,在完成分割和配准研究的基础上,结合金属区域和线路区域的物理特征和分布特征,分别设计了金属区域检测算法和线路区域检测算法。对于金属区域,考虑其形变小,分布区域小的特点,设计了基于局部配准的参考检测法,以灰度值为特征,完成配准,使用形态学处理的方法去除干扰,完成缺陷定位。对于线路区域,考虑其形变大,分布范围广的特点,设计了基于线路骨架的非参考检测法,利用开路、短路、缺失、凸起的骨架特征对缺陷进行定位,对干扰进行去除。实验发现各类缺陷的漏检率和误检率均在5%以下,检测时间远低于人眼检测时间,算法具有一定工程应用价值。更多还原

【Abstract】 As electronic products become thinner and lighter,Flexible Printed Circuit(FPC)has been widely used due to its lightness,thinness,and flexibility.China is the most important FPC manufacturing base in the world.The increase of FPC output and the complexity of FPC products make quality control become an important step in FPC production.Manual inspection method cannot meet the requirements of the enterprises.In this thesis,an FPC provided by a FPC manufacturer is studied.The rigid structure and the flexible structure are staggered,the routing of the FPC is complicated.Therefore,the FPC is representative.Based on the research of the literatures about printed circuit board inspection at home and abroad,this thesis refined and studied two key issues involved in FPC detection:FPC image segmentation and FPC image registration.Based on this,a defect inspection algorithm of the FPC is proposed.In this thesis,the detection areas are divided into metal area and line area.According to differrent characteristics of metal area and line area,image segmentation algorithms are designed.For the local threshold segmentation algorithm of line area,twice the maximum line width is determined as the best sampling side length based on the experiment.Then,the FPC image registration is studied.In order to reduce influence of environment and improve the computing efficiency,an efficient FPC target positioning algorithm is designed.Then,an image registration method that based on the centers of the positioning holes is desighned.A preliminary study of the local area registration method based on local features is finished.It is concluded that the FPC image transformation is non-linear,which provides guidelines for the design of the detection algorithm.Finally,considering the physical and distribution characteristics of the metal area and the line area,detection algorithms are designed respectively.For the metal area,considering the small deformation and small area,a reference detection method based on local area registration is designed.The gray value is used to complete the registration,and the morphological processing method is used to remove the interference.For the line area,considering the larger deformation and wide distribution area,a non-reference detection method based on skeleton is designed.Skeleton feature for differrent defects(open,short,projection,nick)are used to locate defects and remove interference.The experiment results show that the algorhithm can potentially be used in industrial production.更多还原