A density-based fuzzy clustering technique for non-destructive detection of defects in materials

In non-destructive testing and evaluation of materials, defects contain visible aggregations of similar levels of brightness with large scale of correlation between them. In most cases, these brightnesses have no notable contrast relative to non-defect counterparts. However, the density and the size of the defect are visually the most notable features. In this paper, we have utilized human conception for classifying defects by the fusion of fuzzy clustering method and fuzzy logic rules based on the density and the size of the defect. The probability of detection and the probability of error are compared with the Bayes classifier. The proposed approach shows that there is less dependency between the variation of density and size of a defect and variations of noise density and distribution. Experimental images from eddy current, ultrasonic and radiography techniques are investigated. It is shown that the new approach reduces the noise and drift, leading to a better detection of defects.