Color Inspection System for Seat Belt Webbing Using Neural Networks

A system has been developed to take the place of experts employed in the visual inspection of difference in color appearance of dyed seat belt webbing The judgement of experts is likely to be affected by their physical conditions, making the acceptance region of difference in color appearance vague

In this paper, three-layer neural networks are used to approximate the nonlinear relationship between color specification values in the L*a*b* uniform color space (published by the Commission Internationale de I'Eclairage) and human color perception

As a result of using neural networks, the system attained a maximum error rate of 0 9% compared with the judgement by experts and was put in practical use