Detecting the weld position is the precondition of seam tracking during an arc welding process, and is crucial to achieve high-quality welds. Visual sensing is one of the most attractive approaches to detect the weld position, which provides valuable information to control the arc following the expected path. The weld edge is main characteristics of the weld position image, here the visual technique for detecting the weld position based on Hough transformation is proposed. The weld position can be detected and recognized through Hough transformation, while the weld edge image is detected using the Canny operators. And Hough transformation is modified based on the actual weld situation. Therefore, the precision of the weld detection can be improved. From the analysis of experiments, it has been proved that the proposed method can detect the width and the center of the weld, and is suitable for the seam tracking system based on the computer vision.