0000000000347464
AUTHOR
Claudine Coulot
Shape Reconstruction of Metallic Objects by Artificial Vision
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 …
Dimensional control of metallic objects by artificial vision: contribution to lighting condition studies
We present an original method that enables real time automatic dimensional control of defects on metallic objects. The method is based on the theoretical and experimental study of light reflection mechanisms that can be seen as a function of the surface roughness and of the light source size. We show that it is possible to reconstruct the profile of an object from a single gray level image. An industrial application enables us to validate our results. Finally, an error computation fixes the limits of our method and enables us to better understand illumination problems.