Weld Defect Detection in Radiography Based on Projection Profile and Rst Invariant by Using Lvq

X-ray radiography is commonly used in (NDT) Non-destructive Testing, for identifying defects in weld. When the X-ray is passed through the weld object, the area where the defects are occurred will be having different intensity profile, than the nearby pixels. Most of the X-ray radiographic images will be having some forms of noise components embedded in it. Median filter is applied for noise removal, followed by gamma correction for image enhancement which made the image more operative. For the segmentation of the weld defect, watershed method is performed. Through watershed segmentation process, the defective regions are segmented out, without oversegmentation problem. Standard derivation and mean of the Projection Profile of the radiographic image along with RST invariants features are used for feature extraction. In this work, we fed the feature extracted to a Learning Vector Quantization (LVQ) for training, with four different output classes, where each class corresponds to different classes or types of weld defects like Cluster Porosity, Slag inclusions, Lack of Penetration (LOP) and Burn-Through. The result shows that the proposed system is highly efficient in classifying different types of weld defects.