Automated visual inspection using syntactic representation

A knowledge-based automated visual inspection system is described which operates in two modes: learning and production. In the learning mode either a CAD (computer-aided design) file of the components or a standard reference image is converted to an implicit grammar. The grammar, which is a syntactical description of the reference image, is stored in the knowledge base. For the production mode, grammar rules about the inspection of the image are converted to a parse table. The image to be inspected is captured and digitized. Only the small area of the image in which the next image segments are expected is enhanced, achieving considerable savings in processing time. The general-purpose (predictive) parser analyzes the image, reporting component names, locations, the component identification, and possible defects due to syntactic or semantic errors. Finally, a knowledge-based system characterizes the defects. The operation of the current system is described, and implementation considerations are presented