Point cloud registration for non-destructive, high-resolution coating thickness measurement from 3D scans

Thickness is a key specification property of thermal spray coatings. The ability to measure coating thickness nondestructively, in situ, and with high spatial resolution is thus vital for applications like process monitoring, quality assurance, and research. To this end, one of few promising approaches is 3D scanning. By capturing the workpiece before and after coating, thickness can be inferred from the difference in geometry. However, a major challenge stems from the relatively low thickness of typical coatings, which necessitates an exceptionally precise surface registration. We present a complete algorithmic pipeline for non-destructive thickness measurement using optical 3D scanners, suitable for thermal spraying. At the core, we propose an alignment algorithm that registers both scans to an a priori known CAD model. This indirect approach eliminates the need for post-processing of the measured point clouds. A novel weighting scheme accounts for local mismatches in point density of the scans, enabling high accuracy even on incomplete scans. We demonstrate the method on several applications, like spray spot analysis and quality assessment of a production part, and validate the accuracy of the method. Additionally, we provide detailed analyses of convergence properties and the effects of measurement noise, point density, and model mismatch.