Surface data imputation with stochastic processes

TL;DR

This paper introduces a stochastic process-based method using Gaussian processes for surface data imputation, effectively handling spurious measurements by preserving surface characteristics and integrating prior knowledge.

Contribution

It presents a novel application of Gaussian processes for surface data imputation, improving accuracy by incorporating surface knowledge and handling non-stationary features.

Findings

The method accurately imputes missing surface data while maintaining surface features.

Gaussian process models outperform traditional imputation methods in preserving surface characteristics.

Application to real-world data demonstrates practical effectiveness.

Abstract

Spurious measurements frequently occur in surface data from technical components. Excluding or ignoring these spurious points may lead to incorrect surface characterization if these points inherit features of the surface. Therefore, data imputation must be applied to ensure that the estimated data points at spurious measurements do not deviate strongly from the true surface and its characteristics. Traditional surface data imputation methods rely on simple assumptions and ignore existing knowledge of the surface, resulting in suboptimal estimates. In this paper, we propose the use of stochastic processes for data imputation. This approach, which originates from surface texture simulation, allows a straightforward integration of a priori knowledge. We employ Gaussian processes with both stationary and non-stationary covariance structures to address missing values in surface data. In addition, we apply the method to a real-world scenario in which a spurious turned profile is obtained from an actual measurement. Our results demonstrate that the proposed method fills the missing values by maintaining the surface characteristics, particularly when surface features are missing.