In this paper the method by which an autonomous guided vehicle tracks a white-line-course using image processing is described. There are three peculiarities in this method. The first is that the input images of the white-line-course are all segmented and are treated by serial pattern numbers. The second is that each input image has the output values of the steering calculated in advance and all the output tables according to the input images are stacked in the data files. The third is that the path is approximately tracked by connecting the several arcs. We confirmed the effects of this method using simulator which replicated the motion of an autonomous guided vehicle.