This article describes a procedure for adjusting sampling locations in inspected datasets of railway track geometry. These sets are spatially discretized and their sampling locations change with each measurement, meaning that variations in track geometry cannot be simply computed as the difference between two datasets. The difficulty of the adjustment arises from the fact that some sampling intervals locally shorten and lengthen. In the proposed procedure, the adjustment is formulated into an optimization form, and the one unknown parameter involved in the form is stochastically identified. Consequently, the variation datasets are obtained and they demonstrate unknown properties.