Defect Detection on an Inhomogeneous Structured Surface

Abstract

The detection of different defects on a metallic surface is only one part of the various industrial inspection tasks. The recognition of defects (damages) on a structured surface, e.g., wood surface, is a task no less important. The used technique must ensure an accurate recognition of all existing defects independent of the surface structure and brightness variations. For physical reasons, the defects present on a wooden surface (cracks, holes, knots, etc.) feature very sharp edges, but without bulges. The utilization of the SDD algorithm is therefore not entirely appropriate for the recognition of such defects. This technique requires that a target object always has two edges with certain characteristics. This, however, does not necessarily apply for the defects on a wooden surface. As has been already shown in Chap. 1, the known structural analysis methods turn out to be formalistic, complex, and hardly effective when the inspection of naturally structured surfaces such as wood is concerned. One needs other techniques acting quickly and in a robust and adaptive way. This challenge can be met if the defect is first separated from its surrounding and then analysed. For example, this can be done using the contour tracing segmentation, the so-called blob analysis [12]. Here, the detection of edges of an object also plays a major role.