Background
The technical textile sector plays an important role in the global textile industry; in 2018, the Belgian textile sector realized a turnover of €6.1 billion, of which €2.2 billion was realized by the technical textiles segment, with 7,600 employees. As competition from lower-wage countries increases, it becomes more and more important for Belgian textile producers and finishers to consistently guarantee a high-quality product. For bare, woven textiles, inspection has historically been performed by people, watching rolls of textile pass over a light table, and manually flagging imperfections. The thickness of coated textile was typically determined via radioactive sources. As both methods present problems, (inspection by humans is time consuming and imprecise, inspection by radioactive sources is being legally phased out) a transition to automated inspection is taking place. This typically makes use of different cameras, coupled with libraries of textile defects, to alert the user when something has gone wrong. The project Inspect 4.0 seeks to help realize this transition, by illuminating and illustrating the benefits of such systems.
Aim of the project
The aim of Inspect 4.0 is to demonstrate how machine learning and machine vision are combined to produce flexible and accurate quality inspection systems, which can be deployed in a range of textile manufacturing setups. In Inspect 4.0 machine vision is seen as a combination of camera systems that use the wavelength spectrum of electromagnetic waves between the UV and long wave infrared spectrum, (250 nm – 140 μm) complemented with extensive data analysis. The combination of different camera systems can be used to replace and/or enhance existing quality inspection systems, easing the transition to more automated textile inspection. The data analysis will be used for a machine learning system allowing for the detection of errors. Once the illustrative system is set up, investigations will be made into the possibility of incorporating the machine into a broader predictive maintenance framework.
Project consortium
The consortium consists of Antwerp University (UAntwerp) and Centexbel. UAntwerp will focus on the camera systems, while Centexbel will focus on the textiles, both coated and uncoated.
The User Committee consists of all related companies: coaters, weaver, machine builders, etc.
Code examples and training data
PyTorch transfer learning: Google Collab & Data
Bachelor theses
- Yorne Van Praet & Gilles Vanlommel - Visual Quality Inspection with multiple types of cameras on textiles using machine learning (op aanvraag)
-
Julie Gervais & Ward Van Gils - Rol-naar-rol systeem voor textielinspectie (op aanvraag)
-
Brent Cleys & Rhys Evans - Visual Defect inspection of textiles on multiple spectra using machine learning (op aanvraag)
Master theses
- Jordi Muller - Closed Loop testopstelling voor kwaliteitscontrole textiel (op aanvraag)
- Senne Daniëls - Obtaining Consistent Images on a Variable-Speed Conveyor Belt using a Line Scan Camera (op aanvraag)
- Gilles Vanlommel & Yorne Van Praet - Analysis on Textile Samples using Multiple Machine Learning Algoritms (op aanvraag)
- Brent Aernouts - Effect of Light Sources on Textile Quality Inspection (2024)
Presentations
Project reports
