一种新型的PCB图像快速配准算法

PCB图像配准是进行自动光学检测的关键步骤。PCB中往往存在许多相似图形和区域,一般特征点提取和匹配方法效率低,且容易产生误匹配。提出一种以PCB图像中特定几何图形的中心作为特征点,基于相似三角形约束的快速配准算法。提取实测图和标准图中圆形和方形的中心点集进行DT剖分,找出两幅图三角网中的相似三角形集,再对相似三角形的中心点集进行二次剖分和比对,以增强匹配的可靠性。实验证明:该方法计算速度快、匹配正确率高,且能得到均匀分布的正确匹配点。

Fast Image Registration Algorithm for PCB Images

PCB(Printed Circuit Board) image registration is a key step in automatic optical inspection.There are many similar graphics and area in PCB.The general method of extracting and matching of feature points is low efficiency and easy to produce false match.A fast registration algorithm based on the similarity triangle constraint was proposed,which takes specific geometric center as the feature points.The sets of centers of round and square were Delaunay triangulated.Similar triangles were found in two triangulation networks and implemented another subdivision and comparison to enhance reliability.Experiments show that the method is rigid and has higher accuracy,which can get the correct matching point of uniform distribution.