Assessment of surface topography modifications through feature-based registration of areal topography data

Surface topography modifications due to wear or other factors are usually investigated by visual and microscopic inspection, and—when quantitative assessment is required—through the computation of surface texture parameters. However, the current state-of-the-art areal topography measuring instruments produce detailed, areal reconstructions of surface topography which, in principle, may allow accurate comparison of the individual topographic formations before and after the modification event. The main obstacle to such an approach is registration, i.e. being able to accurately relocate the two topography datasets (measured before and after modification) in the same coordinate system. The challenge is related to the measurements being performed in independent coordinate systems, and on a surface which, having undergone modifications, may not feature easily-identifiable landmarks suitable for alignment. In this work, an algorithmic registration solution is proposed, based on the automated identification and alignment of matching topographic features. A shape descriptor (adapted from the scale invariant feature transform) is used to identify landmarks. Pairs of matching landmarks are identified by similarity of shape descriptor values. Registration is implemented by resolving the absolute orientation problem to align matched landmarks. The registration method is validated and discussed through application to simulated and real topographies selected as test cases.