Computer Vision System for Printed Circuit Board Inspection

In this paper we present a Computer Vision system for printed circuit board (PCB) automated inspection. In the last years PCB industry has been invested in manufacturing automation improvement. This is known, especially in measurement and inspection field. We can note that the tolerances on PCB assembly become more accurate. With computer hardware and cameras advances, new Computer Vision algorithms should be developed, and applied in industry with low cost. Besides, new visual inspection systems using computers should be implemented to solve smaller tolerance requirements. A PCB consists in a circuit and electronic components assembled in a surface. There are three main process involved in its manufacture, where the inspection is necessary. The main process consists in the printing itself. Another important procedure is the components placement over the PCB surface. And the third is the components soldering. In the proposed Computer Vision PCB Inspection System we consider the first manufacturing stage, i.e., the board printing. We first compare a PCB standard image with a PCB image, using a simple subtraction algorithm that can highlight the main problem-regions. Then we used connection analysis to find fatal and potential errors, like breaks and circuit shorts. In other to develop this methodology in real PCB, we propose to magnify the problem-regions and start to find the errors in a set of PCB sections, which are smaller than the main PCB image. This approach seems to be very effective if applied in a real time inspection system. Therefore, we propose a new algorithm to solve PCB inspection problem, considering its efficiency in reducing the computational time.