An Artificial Neural Network (ANN) with a raw image Fuzzy pre-processing mechanism system for static colored pattern classification is presented. A computational three layered feed-forward network utilizing a non-linear supervised learning paradigm is trained on fuzzily processed chromatic and achromatic pattern values. During training, center and bandwidth parameters for the tunning of antecedents in Fuzzy rules, corresponding to perceived opponent color categories, are reinforced. These tunned rules are subsequently used to pre-process raw bit test patterns before automatic ANN categorization. By adjusting the opponent primary pairs using the proposed approximate reasoning methodology in conjunction with the ANN, partial human-like visual perception characteristics (primarily color constancy, shape constancy and limited size constancy) are achieved. A particular test bed application has been chosen to demonstrate the usefulness of this system in industrial environments, namely, an automatic visual inspection machine for mounted SMT (Surface Mount Technology) PCB's (Printed Circuit Boards). In this particular application grey-scale inspection proved ineffective due to similar tone scale values of PCBs and some miniature components. Part existence, orientation and correct terminal soldering inspection and classification are being performed under real-time, and real environmental constraints with high hit rates, and low system training trials.