SEMICONDUCTOR DEFECT CLASSIFICATION

Automated inspection has become a vital part of
quality control during semiconductor wafer production. Current
processes are focussed on finding defects via variation from a
‘golden’ image using pixel to pixel comparisons or utilization of
opaque neural network-based approaches. We present a novel
approach which uses the Bag of Visual Words technique to
determine local features that correspond to specific defects within
a wafer image, known as a custom vocabulary, as a way to begin
creation of a more transparent system for automated defect
detection and classification. We demonstrate that the custom
vocabularies, combined with machine learning algorithms, result
in high performance accuracies with efficient computational runtimes.