Disaster Avoidance in Industries through Weld Flaw detection from Radiographic Weld Images using Radon Transform and Improved Fuzzy C-Means Clustering7

The weld flaw detection using image processing techniques helps in avoiding huge disaster in industries such as petrochemical, power generation and nuclear power plant, since this automated defect detection is more reliable and faster than other techniques. This paper proposes a new advanced Fuzzy c-Means (FCM) clustering for the detection of weld defect in radiographic images which arise due to various reasons which affect adversely at the time of welding process. The welding defect such as Burn-through, Lack of Penetration and Slag Inclusion can lead to disaster if undetected at the initial stage of installation of the machine or pipeline etc. Feature extraction is the process of detecting and isolating the distinguishable features for pattern identification which plays a major role in the success of pattern recognition. This paper adopted a new method of extracting feature vector by performing Radon transform for a single angle along the major orientation axis of the weld defect. This paper also proposes a new improved FCM clustering where the objective function is minimised in less number of iterations when compared with ordinary FCM and Hard c-Means. The implementation and result analysis of clustering shows that a better veracity is achieved in less number of iterations with the proposed advanced FCM with Radon transform based feature extraction when compared to Zernike moment based feature extraction.