Robust reference algorithms and softgauges for data fusion-based methods of measured data on 3D objects

For improving manufacturing processes trough quality assessment, measurement systems can be integrated into machines. Data fusion methods, which combine measurements from multiple sensors, can enhance accuracy, speed, and robustness, allowing for larger measuring ranges and higher resolutions. These improvements are particularly valuable for characterising complex shapes and overcoming the limitations of optical measuring systems, which are fast but restricted by their field of view.

The verification of data fusion software is critical to ensure its accuracy and reliability. In metrology, software verification is carried out using softgauges (i.e., reference software or data). However, data fusion methods lack softgauges for verification.

In this thesis, a novel data fusion framework is established to serve as a basis for the development of the reference data generator and software in dimensional metrology. The framework covers registration and fusion processes for heterogeneous sensors and complex shapes while handling groupwise data fusion, anisotropic uncertainty modelling, outliers rejection, and regularisation from prior registration.

The reference software is developed and implemented for Machine Vision Systems (MVS). This software is verified on point clouds from a new reference data generator for unbiased evaluation. The relevance of the data fusion software in real-world situations was also demonstrated with complex shapes measurements using an in-house developed MVS.

Data fusion, geometry processing, metrology, machine vision, softgauges