Shape Reconstruction of Metallic Objects by Artificial Vision

Abstract

Quality control in industrial applications has greatly benefited from the development of tools like artificial vision. The artificial vision approach of quality control has the advantage of being able to perform both non-destructive and contactless controls as well as to work mostly in real time. However, in the case of non planar objects to be controlled, dimensional control may become hard to realize. Indeed, due to the bidimensional aspect of the images, a loss of information is unavoidable. Consequently, obtaining depth information from a planar image is not obvious. Most of the existing methods use several cameras¹ or several light sources², this leads to an increase in the processing time and the installation cost. Other geometrical defect measurement methods require information contained on a 3D grid of the object to be controlled³.