Linking Feature Lines on 3D Triangle Meshes with Artificial Potential Fields

We propose artificial potential fields as a support theory for a feature linking algorithm. This algorithm operates on 3D triangle meshes derived from multiple range scans of an object, and the features of interest are curvature extrema on the object's surface. A problem that arises with detecting these features is that results from standard algorithms are often incomplete in that feature lines are broken and discontinuous. Our novel linking algorithm closes these broken feature lines to form a more complete feature description. The main contribution of this algorithm is the use of artificial potential fields to govern the linking process. In this paper, we discuss the feature detection process itself and then define the linking procedure in the context of potential fields. We present results for both synthetic and scanned models.