Surface determination algorithm for accurate XCT bidirectional length measurements

Abstract

Surface determination plays an important role in XCT bi-directional length measurement, however, its effect on the measurement results is often overlooked or hidden by other error sources. Most of the published research in dimensional field used the Canny algorithm or the surface determination module in VGStudio. Both of them require input from the operator that can also affect the accuracy of the measurements. Alternatively, the marker-controlled watershed (MCW) algorithm has been proven to avoid the latter issue, however, there is no systematic study that evaluated the surface determination algorithm's effect on the accuracy of bi-directional length measurements. In this study a two-sphere reference sample was measured using an XCT scanner and, with the aid of simulations, the effect of the three surface determination methods on bi-directional length measurements was comprehensively studied.The results show that in the presence of ‘streak’ artefacts, a beam hardening error, if the operator does not set parameters appropriately, Canny and VGStudio implementations lead to either loss of surface or large errors, whereas MCW avoids this issue demonstrating its process automation ability. Nevertheless, with voxel calibration, beam hardening correction and data manipulation, MCW and Canny algorithms enable accurate sphere radius measurements (bi-directional measurements), comparable to the accuracy of an industrial tactile coordinate measuring machine.