When analyzing multiple mesh models, an important first step is to find the rigid transform that best aligns them. Many mesh alignment techniques exist, but because they treat input meshes as a collection of surface samples, they tend to be sensitive to surface variations between those meshes. We instead wish to align shapes based upon an analysis of mesh topology, thus making it possible to align shapes with similar part structures but varying surface geometry. To this end, we present a novel algorithm that extracts a curve-skeleton from a mesh based on its connectivity, rather than its surface geometry. We show that this method is robust to surface variation and achieves consistent results across shapes in the same object class. We then will present our system for mesh alignment based on our topological analysis of input meshes, and we will show that our technique is able to align shapes with similar part structures but differing surface shapes.

Keywords: Shape analysis, Mesh alignment, Skeleton extraction, Segmentation, Algorithms