Vision Inspection System for MIG Welding Joints Using Different Feature Extraction Methods

In this paper, an efficient technique has been described for inspection of Metal Inert Gas welding (MIG). A machine vision system has been developed for identifying and classifying the surfaces of butt joint as per standard EN25817 in MIG welding.Images of welded surfaces are captured through CCD camera. Then regions of interest are segmented and the average gray levels of the characteristic features of these images are calculated using 2D feature vector and Gaussian distribution based features. Finally, welded joints can be classified into one of the four pre-defined images based on the back propagation neural network. In this work, 80 real samples of images are tested and performance of the vision system is compared with twodifferent feature extractions. vision inspection system using Gaussian based feature extraction method produced 93.75% than 2D feature extraction method which is produced 92.5%.