An Automatic detection method of fluorescent-magnetic particle indications for steel billets and welds using the high sensitive CCD camera and image processing techniques has been studied based on a personal computer. In this paper, the automatic detection system consists of two main processes, the extraction of flaw indications and their quantitatively sizing. The flaw indications in the billets and welds are first extracted from digitized images by means of the smoothing, the line segment extraction with the oriented templates and then the component labeling algorithms. As a result of application of these image processing techniques to detection of flaws, the effect of background noise and unwanted indications has been reduced and besides the image quality of desired flaw indications has been enhanced.
For the extracted indications of fatigue cracks with various depths in weld, the relationship between the crack depth and the degree of brightness of the corresponding indication has been experimentally investigated, and the results show that there is possibility to evaluate roughly the size of cracks by the brightness level of extracted indications.