WenDeKI

Surface inspection of components in series production - generative AI for training camera-based inspection systems

Project description:

In series production, for example in the metal and plastics processing industry, the reliable and complete inspection of components is an important factor for quality assurance. Camera-based inspection systems based on AI models offer an efficient and cost-effective solution for this. However, the detection of defects where only a few sample parts are available poses a challenge.

The project aims to solve this challenge by developing a generative AI model that can synthesise the image data of the defective parts required for training. The generative AI model is trained on the basis of a small number of real images of defect parts. It learns to recognise the features of the defects and - this is the decisive step - to generate new, plausible images of defects. This expands the inspection model's training data set and improves the quality of recognition and classification.

In order to test and ensure the transferability of the approach to different industries and use cases, flexible optical inspection systems are used to collect the image data. In particular, a tunnel inspection system is being developed that will be used in two different applications. The approach will be validated in different application areas close to industry, so that a direct transfer is very likely after successful project completion.

Consortium:

SOTEC GmbH & Co KG;
Fraunhofer Institute for Physical Measurement Techniques IPM;
Quittenbaum GmbH;
PCM.de GmbH;
BIA Kunststoff- und Galvanotechnik GmbH & Co. KG

Duration:

February 2025 - January 2028

Budget:

Total funding: € 3.1 million
Funding amount: € 1.7 million