Classification learning method for PCB insertion holes based on shape context

With the increasing demand for Printed Circuit Board (PCB) in the field of modern industrial, it is a trend to use machines to plug in components automatically. In the process of inserting components, classification learning of insertion holes is an important part. Traditional classification learning is completed by means of manual marking, which is not only inefficient but also error-prone. In view of these problems, this paper proposes a classification learning method for PCB insertion holes based on shape context. The core of this method is traversing of the insertion hole contours without having to traverse the whole image. The type and position of insertion holes are determined by firstly extracting the contour points of PCB insertion holes and then matching computed shape context descriptors with template shape context descriptors. Experiments show that this method can accomplish the insertion hole classification learning accurately and efficiently. And this method meets the requirement of actual production.