Abstract

Given the many successes of the machine vision industry and the pervasive, uncritical belief that manual inspection systems should be automated, most engineers and managers are surprised to find that large quality improvements can be realized by optimizing human-based inspection systems. In a recent automotive paint inspection research program large, statistically-reliable improvements in human inspection performance were demonstrated in-plant based on warranty return data alone. The payback time for a complete retrofit of the inspection lighting was under six months and the time to recover the total R&D expenditure was under one month. In this paper the inherent characteristics of inspection for appearance defects on Class A automotive surfaces are described and it is argued that humans are well suited for this inspection task.