Investigation of a fuzzy grammar for automated visual inspection

Abstract

Automated visual inspection involves complex image processing and analysis. Representation of images in syntactic form facilitates parsing of images so that defects can be identified as syntactical errors in the same way that programming errors are detected by a language parser. The syntactic approach for image processing uses image grammars to describe images and a parser to recognize images. A problem with the syntactic approach is that the conventional grammars are too formal to handle uncertainty caused by various noise and pattern distortions in images.

In this dissertation, a fuzzy grammar based on fuzzy set theory is investigated for automated visual inspection. A fuzzy context-free grammar is defined and applied for inspection of VLSI circuit patterns. This dissertation contributes to syntactic analysis of images by enhancing descriptive power of grammars to handle vague descriptions of images.