How to carry out visual inspection more efficiently and more effectively: the characterisation and evaluation of aesthetic anomalies

For some companies, visual inspection becomes an essential step when seeking to improve the quality of their products. The aim of this control is to be sure of the perceived quality of the product, which often goes well beyond the quality expected by the customer. For this type of control, the controller should be able to detect any anomaly on a product, characterise this anomaly and then evaluate it in order to decide if the product should be accepted or rejected. This paper describes how this characterisation can be carried out and, more specifically, how to measure the impact of the local environment of an anomaly on the perceived quality of the product. It also details how an evaluation can be carried out by using a neural network. Finally, the paper details how the knowledge about the visual inspection of aesthetic anomalies may be made explicit to be shared by controllers more easily.