Abstract: To solve the problem of area classification in defect detection of flexible printed circuit,a method for extracting the surface contour of the flexible circuit board and identify specific areas was proposed.To accurately extract the overall outline of Flexible Printed Circuit (FPC) image,the area growth method was used to extract the image outline for the problem of uneven illumination with many spots and impurities on FPC surface image,and the central neighborhood grayscale method was used to eliminate the problem of undergrowth.To effectively identify the contour types of image different regions,a two-way difference method was used to calculate the discrete curvature of the image contour,and Earth Move Distance (EMD) was used to evaluate the difference between the curvature features and the template contour,so as to realize the recognition of the specific area.

Key words: differential geometry, flexible printed circuit, image contour extraction, specific area recognition

摘要： 目前柔性电路板(FPC)的表面缺陷检测方案大多缺少对轮廓进行分类这一步骤,而直接对特定区域如镀通孔、线路部位进行缺陷检测,难以直接应用到实际生产中。为解决这一问题,提出一种能有效提取柔性电路板表面轮廓并进行特定区域识别分类的方法。在该方法中,为精确地提取FPC图像整体轮廓,并有效过滤掉图像前景和背景相互夹杂的部分,针对FPC表面图像光照不均匀以及斑点杂质较多的问题,采用区域生长法提取图像轮廓,并利用中心邻域灰度法来消除欠生长的问题;为有效地识别图像不同区域的轮廓类型,利用双向差分法来计算图像轮廓的离散曲率,利用陆地移动距离(EMD)来评价各个轮廓的曲率特征与模板轮廓的区别,实现了FPC图像特定区域的识别。

关键词: 微分几何, 柔性电路板, 图像轮廓提取, 特定区域识别