Purpose - In the last years, artificial vision systems have widely spread among quality control processes in many industries. In this paper we present an artificial vision system that automatically inspects sachets of surgical material in real-time. These sachets are difficult to inspect by an artificial vision system due to high variability in their visual aspect and the material they are made of (aluminium and plastic) which can provoke a lot of reflections and shadows. Design/methodology/approach - Due to variability of sachets and reflections in aluminium and plastic, the design of a good illumination system is very important for the success of the system. For that reason, we have used two sources of illumination, a dark field illumination and a diffuse frontal one, to avoid reflections and enhance important visual features. Moreover, the output of the system is not only the classification of the sachets as correct or defective, but it also identifies the sachet defect among the fifteen possible ones, a useful feature for just in time production. The proposed system is based on a PC modular and scalable architecture. Findings - The system is currently working in actual production fulfilling the problem requirements with respect to false positive and false negative, in spite of the high variability of the product. Originality/value - The system proposes two different kind of illumination systems to inspect difficult materials, such as plastic and aluminium, achieving good results. The proposed architecture of the system is modular and scalable and allows to increase computational power for visual tasks. © Emerald Group Publishing Limited.
- Quality control
- Visual perception