A fiducial-aided data fusion method for the measurement of multiscale complex surfaces

Abstract

Multiscale complex surfaces possessing high form accuracy and geometric complexity are widely used for various applications in fields such as telecommunications and biomedicines. Despite the development of multi-sensor technology, the stringent requirements of form accuracy and surface finish still present many challenges in their measurement and characterization. This paper presents a fiducial-aided data fusion method (FADFM), which attempts to address the challenge in modelling and fusion of the datasets from multi-scale complex surfaces. The FADFM firstly makes use of fiducials, such as standard spheres, as reference data to form a fiducial-aided computer-aided design (FA-CAD) of the multiscale complex surface so that the established intrinsic surface feature can be used to carry out the surface registration. A scatter-searching algorithm is employed to solve the nonlinear optimization problem, which attempts to find the global minimum of the transformation parameters in the transforming positions of the fiducials. Hence, a fused surface model is developed which takes into account both fitted surface residuals and fitted fiducial residuals based on Gaussian process modelling. The results of the simulation and measurement experiments show that the uncertainty of the proposed method was up to 3.97 x 10 -5 μm based on a surface with zero form error. In addition, there is a 72.5% decrease of the measurement uncertainty as compared with each individual sensor value and there is an improvement of more than 36.1% as compared with the Gaussian process based data fusion technique in terms of rootmean-square (RMS) value. Moreover, the computation time of the fusion process is shortened by about 16.7%. The proposed method achieves final measuring results with better metrological quality than that obtained from each individual dataset, and it possesses the capability of reducing the measurement uncertainty and computational cost.