Detection and Classification of Defects in PCB Images by Manipulation with Color and Machine Learning

Thinning and filling algorithms are used to highlight and label four types of defects: short, open, absent, and additional. The overlay algorithm, using a combination of the defect scheme and the correct skeleton, makes it possible to find different types of defects by assigning them different colors. Color intensity is taken as input for machine learning, represented by an artificial neural network.