ICP fitness analysis for 3D scan data matching

Abstract

Full automation of the registration of 3D scan data is, in general, still an unsolved problem. If supplementary data is provided, either by human assistance, or by additional devices, the registration process can be completed. We have analyzed under which conditions supplementary data is required, where the conditions are specified as subsets of configuration space. As a matching tool an ICP-based (iterative closest point) algorithm was deployed. A point in configuration space represents the amount of translation and rotation needed to obtain a match between 3D scans of two partial overlapping surfaces. As a function of the coordinates in configuration space, we determine the successfulness of the algorithm and hence whether additional information is required. Our shape matching method does not rely on pre-computation of surface invariants, nor on the identification of shape features. The possible application of the analysis results for practical 3D scanning purposes are described.