Matching point clouds: limits and possibilities.

Abstract

In computer-aided production of fixed dental restorations, the process chain always starts with digitizing, independent of the type of further (data) processing, the material used, and the kind of restoration to be produced. The quality of the digitized data, followed by the influences of further data processing and the production parameters, decisively influence the fitting accuracy of the dental restoration to be fabricated. The accuracy with which individually measured 3D data sets in the form of point clouds can be matched for further processing in one common system of coordinates was the object of the present study. Casts of the maxilla and mandible were digitized in several partial measurements comprising two to three teeth in each case, using an optical three-coordinate measuring system. The individual segments were sequentially aligned to surfaces that were created on the basis of partial point clouds. The mean deviation between surfaces and point clouds was between 1.90 microns and 18.24 microns. The accuracy of the alignment was determined by the RMS (root mean square) error, and was on average 14.2 microns (SD 7 microns) for the maxilla and 17.2 microns (SD 9.4 microns) for the mandible. Combining a larger number of smaller segments did not improve the result, since the errors of the individual registrations are summed in sequential matching. In this study, the errors arising in matching are not negligible and can possibly negatively influence the quality (fitting accuracy) of the restoration produced on the basis of the matched data records.