A cast system for visual fault detection research

Abstract

Building a theoretical model, an analytical model and an equivalent neuronal net to solve the fault detection problem, permits to define a formal frame work, from CAST, CAD, and CAM techniques, which permit to try several resolution methods and to make an appropriate assessment of the results. This model, first intended to detect faults in leather, can be generalized to a large number of problems related to quality control in automatic chain production.