- fr
- en
Jacques Harvent
Abstract
Shape measurement using a multiple-view Digital Image Correlation (DIC) method: application to the inspection of large-scale aerospace structures
Jacques Harvent - 5 November 2010
In 2006 the Ecole des Mines d'Albi and the LAAS-CNRS initiated a collaboration with AIRBUS Toulouse and EADS-IW for the development of a computer-vision-based system for the inspection of aeronautic parts (fuselage parts, metallic or composite aeronautical panels, etc.) in order to detect shape defects (shape deviation with regard to the desired overall shape corresponding to the CAD model). The system is composed of several cameras (at least four) giving the capacity to inspect large parts. The cameras are fixed on a rigid structure and the parts to be inspected are positioned in front of the system. Several images are acquired synchronously and they are used to reconstruct the 3D model of the part. This thesis focuses on different strategies that can be developed to manage a multiple-camera system. The different steps of the digitization process are presented, namely: multiple-view digital image correlation (a multipleimage DIC criterion well suited to the multiple views context is addressed), 3D cloud stitching, calibration assistance unit. Using more than two cameras has brought multiple benefits. It allows digitizing large aeronautic parts (several m²), provides the whole shape of an object in a one-shot acquisition, improves the accuracy thanks to data redundancy, and permits to avoid specular reflections on non-lambertian surfaces. Our algorithms have been evaluated through synthetic images as well as through the comparison with measurements acquired by different commercial digitizing systems.
Plan du site