A novel background subtraction technique based on gray-scale morphology for weld defect detection

Optical inspection is a non-destructive quality monitoring technique to detect defects in manufactured parts. Automating the defect detection, by application of image processing, prevents the presence of human operators making the inspection more reliable, reproducible and faster. In this paper, a background subtraction technique, based on morphological operations, is proposed. The low-computational load associated with the used morphological operations makes this technique more computationally effective than background subtraction techniques such as spline approximation and surface-fitting. The performance of the technique is tested by applying to detect defects in a weld seam with non-uniform intensity distribution where the defects are precisely segmented. The proposed background subtraction technique is generalizable to sheet, surface, or part defect detection in various applications of manufacturing.