Mining Linguistic Information for Configuring a Visual Surface Inspection System

The configuration of a surface inspection vision system as a complex task, requires mining associations among attributes due to the variability of the surface and the environment in real-time production process. The surface inspection task has to change to deal with different elements such as, wood, stainless steel or paper inspection and, in the case of stainless steel, with reflectance and thickness variations. This work is an approach to mine linguistic information based on the fuzzy transform method, capable of finding associations among features. We are interested in the linguistic form of the observed associations, derived from numerical data, as the configuration task needs such linguistic knowledge. This linguistic knowledge is handled by the dynamic cognitive architecture (ARDIS), here propose, to deal with the system knowledge, represented by means of IF-THEN rules, required to configure a specific inspection system.