The product appearance plays an important role in the perception of quality by the customer. Beyond the features that must be satisfied, now a product must have a flawless appearance. However, there is no perfect surface, because a given level of magnification, a deviation from an ideal surface can always be identified. To detect this deviation and assess its impact on perceived product quality, companies usually set up a visual inspection of the surface appearance of their products. A first PhD thesis was carried out at the Laboratoire SYMME to propose a methodology to reduce the variability generally observed on the results of this type of inspection. Our work is in the continuity of this thesis with the aim to propose methods and tools for the control of three stages of visual inspection of surface appearance: exploration, evaluation and decision. The thesis project carried out under of a European research program INTERREG IV brought together different Universities and Companies. The practices of corporate partners have brought a testing field for the proposed researches. Based on this observation, we proposed a conceptualization of human visual inspection leading to proposals for methods and tools adapted to the three stages. These proposals were tested in the partner companies to verify their robustness to a variety of industrial situations. For example, we proposed a new test R2&E2 Compliance which measures the variability of a visual inspection and helps to identify possible sources of this variability. In addition to this conceptualization for the development of tools, we list a set of recommendations to be followed by companies for a better exploration of anomalies. We also propose a set of sensory attributes to characterize, with a view to evaluate any anomaly appearance. Finally, we show how to formalize the expertise process, a controller can evaluate an anomaly appearance and judge its impact on the perceived quality of the product.
