PCB inspection based on a variant of the n-tuple technique

Proposes an alternative approach based on a variant of the n-tuple technique to detect small defects in electronic printed circuit boards. The technique operates on a list of nine attributes formed in a 3 by 3 kernel. The comparison of corresponding frequencies of occurrence of these selected attributes recorded in the faulty and the fault-free images, provides an adequate means to categorize the defect. The results obtained during the study, show that the frequency distribution of the nine attributes gives an estimation of the size and shape of the faults. This suggests the use of this rapid technique in a preprocessing stage within a complete autonomous PCB inspection system.<
>