The use of machine vision technology in the welding process has been developed along with the need for more consistent and more qualified welding results. The application of machine vision to perform object analysis without direct contact to the object itself making the whole processes fast while maintaining the accuracy. In this study, some image processing algorithms are combined which include : image enhacement (grayscaling and median filtering), edge detection, binarization, and Hough Transform for line detection. The candidate of welding tracks are produced through this standard algorithm. However, afterward, some modifications to the Hough transforms are carried to enable finding the actual welding tracks. The implemented method is capable of detecting the actual welding tracks and their appropriate welding path with accuracy not more than 0.9 mm.