Estimation of ionic contamination in printed circuit boards from shape and color features of dendritic patterns

Ionic contamination is one of the significant causes of failure in printed circuit boards. There are several ionic contamination measurement methods for printed circuit boards using specialized equipment, but a possible alternative method may be constructed based on visual analysis. Using the water drop test, dendrites and other contamination products produced during electrochemical migration may easily be observed. This paper proposes a novel way to automatically estimate ionic contamination level in a printed circuit board using shape and color features of the failure pattern. In this study, optical microscope images of the failure pattern formation were obtained by performing the water drop test using NaCl-contaminated water at low, medium, and high concentrations. Shape features describing the isolated dendrite pattern, such as fractal dimension, convexity, solidity, eccentricity and branch density, were obtained. Color features in HSV color space were also extracted. K-Nearest Neighbor was used for classification. While the use of shape features was effective to detect low contamination levels, it had limited effectivity at higher levels due to the obscuration of branches caused by bubbles and overlapping, along with the naturally high variation of the observed shapes. The addition of color features captures the visual changes both in the dendrite and the precipitate associated with contamination and improves the result.