Because welding system is given the look function by laser image sensing, the automation of welding system is well improved. But the two problems are image processing deviation under welding interference and adjusting deviation of machine set tracking. Targeting at these problems, the weld seam tracking (WST) system is designed. Some mathematics are brought forward, including the log grey transformation, Laplace edge detection, extracting center route of laser strip and detecting the characteristic points of weld seam image. The dynamic trend of WST is controlled by the current seam deviation calculated by seam characteristic points. The subsection control strategy and the decision algorithm of WST stability in abnormal circumstances are used. The WST experimental results show that tracking S shape weld seam is delivered effectively with high precision based on the proposed weld seam tracking, can meet welding technology requirement and provide experimental data for manufacturing WST system.