An object-based approach for bare PCB inspection

摘要

A novel object-based approach for defect inspection and classification on a bare printed circuit hoard (PCB) is presented. In this technique, the reference image is first segmented to generate a template tree consisting of a set of objects that describe the shape characteristics of the reference image. Three reference images are then introduced to locate the defect-objects. All kinds of defects involved in a PCB can be detected through the comparison between the inspected image and the reference images. The template tree helps to produce the reference images and classify the defect-objects. The process of classification is simplified according to the characteristics of the location of the defect-objects. Since the segmentation in the reference image is processed in advance, the computational complexity on inspection is much reduced. Experimental results show that the total inspection and classification time is 0.05 sec for an image with pixels of 640 × 480 when executed on a Pentium-III computer.