Pinhole detection in steel slab images using machine vision

AbstractIn this paper, we introduce an image-processing algorithm that detects pinholes in steel slabs. Because the pin-hole is relatively small and similar to noise components, it is difficult to distinguish pinhole from a steel slab surface. To increase detection accuracy, we propose a new algorithm for pinhole detection. We applied Gabor filters to extract defect candidates and subsequently developed a morphological score function to evaluate the specific features of each pinholes. The performance of the proposed algorithm was evaluated on 2442 subimages; 47 of these images had pin-holes, and the rest of the images were free of defects. The proposed algorithm has 100% accuracy for the detection of pinhole images and 97.4% accuracy for the detection of defect-free subimages.