Development of a data fusion pipeline for all-optical dimensional measurement
Title: Development of a data fusion pipeline for all-optical dimensional measurement
Start: October 2020
Student: Zhongyi (Michael) Zhang
Supervisors: Richard Leach, Sofia Catalucci, Adam Thompson
Data fusion allows for comprehensive measurement of an object by fusing data collected by multiple sensors and has been widely used in dimensional metrology in the past decade. For dimensional measurement, data fusion allows for the evaluation on the shape of the measured object using information and data from a variety of sensos, to provide improved accuracy and coverage compared to single-sensor measurement system.
This project is aimed at developing data fusion solutions for form and surface texture metrology based on machine learning. To test the developed algorithms, multiple measurement technologies will be employed, including fringe projection and coherence scanning interferometry. Test datasets will be collected from artefacts with various geometrical complexities. The project will involve testing existing algorithms, using the data generated during the project to understand their advantages and disadvantages in terms of alignment accuracy and computational efficiency for measured objects with highly complex geometrical features. Additionally, new algorithms will then be proposed to combine the advantages of the existing algorithms with machine learning techniques, such as clustering, grouping and regression.
The data used throughout this project will be collected using the commercial ATOS Core 300 system (GOM, Zeiss) and the in-house JABEH multisensor measurement system. Initial work has focussed on attempts to fuse the data from both systems with the alignment algorithms embedded in open-source point cloud analysis software, CloudCompare. To test and develop algorithms, Python and other programming environments will be used in the later stages of this project.
Figure 1: Data collected by ATOS (pink) and by JABEH (green), coarsely aligned in CloudCompare.