The change of the gap width is one of the important factors that affect the quality of K-TIG welding, and it is difficult to keep the gap consistent in the welding of large workpieces. Vision technology is a reliable method of weld pool monitoring. To avoid the influence of the strong arc of K-TIG welding, this paper designed a real-time HDR system and based on which, a weld pool and keyhole geometric feature extraction algorithm was proposed. The designed HDR system includes an automatic exposure algorithm which is suitable for HDR and an improved HDR fusion algorithm which uses exposure instead of exposure time for HDR fusion. Finally, the dynamic behavior of the weld pool and weld forming with gradually changing gap were analyzed. And conclude that under the same gap, whether the gap changed from small to large or from large to small suddenly, the weld pool will collapse. In the case of the same gap change rate, the weld forming of the welding direction from a large gap to a small gap is much worse than that from a small gap to a large gap. Therefore, it is necessary to avoid a welding direction from a large gap to a narrow gap.