In the frame of the ULB spin-off project named RePRO glass we are developing a new optical technology aiming to perform quality inspection for car windscreen industrial applications. This technology is basically working as a large scanner: the windscreen is translated by a linear stage system in front of the camera system and shape and thickness maps are retrieved from the captured images of the camera system. In general we have two types of signals (Phase Measuring Deflectometry – PMD and Refractive-PROfilometry-RePRO signals) captured by the cameras of the camera system: one signal is a pattern deflected by the windscreen and it is used to retrieve the shape (PMD camera signal) while the second signal is a pattern projected (by a projector system) behind the object and it is used to obtain the thickness information (RePRO camera signal). An alpha version of the RePRO glass prototype has been integrated in the chemical-physics department laboratory of the ULB. The RePRO cameras capture images of the signal through the windscreen and they have sufficient resolution to detect superficial and internal defects of the windscreen. The cosmetic inspection is mandatory during the quality inspection procedure of car windscreen. Depending on the type of defect, a recognition algorithm must be implemented to determine if the defect is superficial (hair, scratch, etc…) or internal (air bubble, trapped contaminating body, etc…). In addition the algorithm should detect if the defect is removable by the intervention of the operator or not. An initial database of the possible defect must be implemented according to the experimental tests. The following tasks must be achieved:

i) Literature review of the methods used to detect and identify cosmetic defects on objects.

ii) Implementation of a C++ solution to process experimental RePRO images.

iii) Implementation of a recognition algorithm based on a specific database to be implemented with the support of experimental data.

iv) Redaction of the relative documentation in the RePRO glass product design document.

Keywords: Image processing, quality, inspection.

Type of work: 10% experimental, 80% coding, 10% documentation.

Contacts:
Dr. Wassilis Tzevelecos wtzevele@ulb.be, Olivier Debeir odebeir@ulb.ac.be

Supervisor:
Ing. Vincent Van Pelt vvpelt@ulb.ac.be