Polarization imaging for 3D inspection of highly reflective metallic objects

In the field of industrial vision, the 3D inspection of highly reflective metallic objects is still a delicate task. The specular reflections prevent the use of a 3D laser scanner, whereas the phase shifting-based 3D systems are more adapted to inspect surfaces with low curvature. This paper deals with a new automated 3D inspection system based on polarization analysis. Studying the state of polarization of the reflected light is very useful for obtaining information on the normals of the surface. An extension of the shape from polarization method from dielectric to metallic surfaces is demonstrated. Then, an improved relaxation algorithm is provided in order to reconstruct the shape from the normal field given by the polarization analysis. Finally, applications to shape defect detection are discussed and the efficiency of the system in discriminating defects on highly reflective metallic objects made by stamping and polishing is presented. The text was submitted by the authors in English.