During high power disk laser welding, the high-speed photography was used to measure the dynamic images of the laser-induced plume at different laser welding speeds. Various plume features (area, height and brightness) were extracted from the images by the color space clustering algorithm. Combined with observation on the surface and the cross sections of welding samples, the effect of welding speed on welding stability was analyzed. From the experimental results, it was found that these features of plume could reflect the welding state. Thus changes of the plume features corresponded to different welding speeds, which was helpful for monitoring the laser welding stability. During high power disk laser welding, the high-speed photography was used to measure the dynamic images of the laser-induced plume at different laser welding speeds. Various plume features (area, height and brightness) were extracted from the images by the color space clustering algorithm. Combined with observation on the surface and the cross sections of welding samples, the effect of welding speed on welding stability was analyzed; the plume features corresponded to different welding speeds, which was helpful for monitoring the laser welding stability.