The project aims to investigate a computer-vision based assistance system to assist human operators to avoid quality relevant manufacturing mistakes.
Manufacture-Assist
Partner Call open until: February 25, 2025
Project Start: March 2025
Process mistakes in mechanical assembly by human operators can lead to costly quality defects. The project focuses on the detection of undesired events in a series of process steps. The specific goals include:
- Establishment of an optimized camera sensor concept, including sensor selection and their positioning with respect to human operators.
- Investigation of optimized feature extraction from time-aligned multi-view videos to identify key spatiotemporal features.
- Development of the process relevant decision-making system; including the exploration of various supervised and unsupervised machine learning approaches.
Expected results
- First measurement-based feasibility results of a multi-view machine learning based assistance system.
- Insights into practical human-in-the-loop AI systems in a manufacturing environment.
- A performance comparison of various machine learning approaches in this context.
