Detection of Circular Defects on Varnished or Painted Surfaces by Image Fusion

Varnishing and painting are quite delicate processes susceptible to failures. This becomes especially critical whenever an aesthetic finish is decisive for the final product. Such is the case, for example, in production of furniture, musical instruments, and paper, among many others. Dozens of different defects with very varied shapes can affect varnished or painted surfaces. However, an important group of them presents a mere circular geometry. Circular defects can have very different sizes and are only partially visible under certain illumination conditions. These characteristics complicate the quality control of varnished or painted surfaces, for which an automated solution does not exist. On the contrary, this work is performed by human operators, which yields to a long, expensive and subjective inspection process. In order to implement an automated surface inspection, different lighting directions and a priori knowledge on the defect symmetry are combined. The image series, resulting of sampling the illumination space, presents the complete information of defects, but distributed on its pictures. To collect all this information, an iterative algorithm based on a multidimensional correlation filter is proposed, which enables a reliable identification of circular defects with arbitrary diameters. Moreover, the successful detection and identification of these defects is not affected by the surface texture