| Zugriffsnummer | 22182 |
| Dokumenttyp | Konferenzartikel |
| Sprache | Englisch |
| Titel | Sphere fitting algorithms with uncertainty estimation |
| Autor(in); Institution |
Danzl, R.; Alicona Imaging GmbH, Graz, AUSTRIA
Helmli, F.; Alicona Imaging GmbH, Graz, AUSTRIA
Krystek, Michael; 5, Fertigungsmesstechnik, PTB-Braunschweig
Neugebauer, Michael; 5.3, Koordinatenmesstechnik, PTB-Braunschweig
|
| Quelle/Jahr | ICPM 2008: Proceedings:(2008) |
| Availability | [CD-ROM] ; file name: 2_0_24.pdf |
| Herausgeber(in) |
Scharff, Peter
|
| ISBN | 978-3-938843-40-6 |
| Verlag | Ilmenau: Techn. Univ. |
| Konferenzangaben | International Conference on Precision Measurement (ICPM), Ilmenau, 08-12, September, 2008, Germany |
| Freie Schlagworte | measurement uncertainty ; mathematical algorithms ; optical sensors |
| Zusammenfassung | A common coordinate measurement task is to measure the form and geometry of small parts. This often involves the fitting of geometric primitives into measured points and to give an estimation of the uncertainty of the measurement. We analyse two algorithms that solve the task of fitting a sphere in the least-squares sense and compare the results of the calculated sphere parameters as well as their standard uncertainties. The first algorithm has been developed in PTB, the second is part of the software used by the optical metrology device InfiniteFocus developed by Alicona. Both algorithms base their estimation on the mathematical analysis of a single dataset. The algorithms only estimate parts of the full measurement uncertainty (e.g. components that are due to form deviations of the object or due to some aspects of the measurement device) but do not include other uncertainty parts such as scaling errors. In order to analyse the two algorithms, experiments have been performed on synthetic and real datasets with different constraints on the input data. Both algorithms leaded to very similar results with absolute differences < 1 nm for all sphere parameters and relative differences for the standard uncertainties that are below 0.2% for 23 out of 25 evaluated datasets. |
