Objective

In one inspection cell of a production line, as in manufacturing of metallic parts, a vision system captures images of the parts to detect possible imperfections. The objective is to make a quality control of the parts. Before the implantation of this inspection cell but, it is required to do a lot of tests and evaluations of the vision system to correctly detect the imperfections. So many times this task can be very difficult to carry out. In that context, our purpose is to develop a graphic system that helps in the evaluation and training of this kind of systems. We want to generate images in an easy and intuitive way that simulates the ones obtained by the real camera on the inspection cell. In this way, the vision system could be tested using our images as inputs instead of the ones obtained by the camera, without the need for installing the system on the production line. This fact will suppose an improvement on the training phase, avoiding the typical drawbacks often involved with this: travels up to the manufacturing system, production line stopping, equipment assembling, etc.; involving a great save of time and resources. To obtain these images, our idea is that the user might be able to virtually simulate the real environment of the inspection cell, with all its elements, like the platform, the different parts, the cameras, the illumination, and the different imperfections over the parts. Finally, once created the desired environment, he should be able to select one of the cameras and obtain the final image. This kind of process has the advantages that allow working with parts not yet manufactured, or test lights and cameras not available physically, for example. Starting from all these ideas, we have developed in this project an initial prototype of the system, mainly based on the realistic simulation of defects over metallic parts, which forms the basis of the system.

 
This project has been done in collaboration with the vision group VICOROB from the IIiA and the Equipe de Visualisation et Rendu Realiste from the University of Limoges (France).