Selection of autofocus algorithms for printed circuit board automated optical inspection system

Abstract

This paper presents an examination study of 11 autofocus algorithms for printed circuit board (PCB) automated optical inspection (AOI). A selection of an optimal algorithm for that application based on some criteria was carried out. Unlike microscopy, PCB optical inspection does not require very high magnification. The object in this work was also different from that of microscopy and thus influenced the image features. We analyzed 47 PCB images, size of 640×480, sequentially captured every 1 mm in the <em>z-</em>direction. This work utilized USB digital microscope, and the magnification was set at ten times. Each algorithm calculated the sharpness values of the image sequences, and the plot of the sharpness profile was created. Moreover, the research also carried out experiments in several strategies, including image resizing and applying the non-local means (NLM) denoising filter to assess the algorithm performance in different situations. The algorithms were examined and ranked based on five criteria, i.e., computation time, full width at half maximum (FWHM), accuracy, number of half maxima, and range. The experimentation results showed that the Brenner gradient worked best for analyzing images both in their original dimension or resized images.