Success Story: AI-powered Optical Sorting Solution for IFSYS

Initial situation: existing sorting solutions do not differentiate sufficiently.

IFSYS, a medium-sized company from Großbardorf, specializes in custom machinery in the field of feeding technology, primarily for the automotive industry. Motivated by the potential of new markets and close contact with agriculture, IFSYS decided to develop an optical sorting machine for hazelnuts. The challenge: existing image analysis algorithms only recognized shape and size, failing to capture the detailed quality criteria of hazelnuts sufficiently. Thus, IFSYS approached us to develop an AI-based solution capable of assessing details such as shell coverage, wrinkles, and size, ensuring precise and reliable sorting that they could redefine and adjust as needed.

Due to the cost pressures in the agricultural sector, automation is indispensable.

Solution: together with the customer, we integrated AI directly into the optical sorting machine.

Close cooperation throughout the project ensured that mechanics and AI were perfectly aligned. What we did:

1. Image capture & processing:
   • Images taken from four perspectives were preprocessed to allow the AI to optimally recognize the relevant nut characteristics.
   • Data augmentation increased the variety of training data.
2. AI training & integration:
   • Development of a deep learning model for precise classification.
   • Easy integration into the machine’s standard PC with quick connection to controls.
3. Smart communication:
   • Aggregated decision logic (post-processing) from four perspectives.
   • Real-time feedback to mechanics to reliably assign each nut to the correct quality category.

“With prognostica, we had a competent partner, especially appreciating their pragmatic approach. Together we were always able to quickly respond to new requirements and find solutions. Ultimately, we achieved more than we initially thought possible.” - IFSYS Integrated Feeding Systems GmbH

Results: enormous efficiency gains open new market potential.

Collaboration with IFSYS led to a successful prototype delivering impressive results:

1. Near-perfect recognition accuracy for various hazelnut classes.
2. Huge efficiency gains and new products: Customer Valeria automated the sorting of their hazelnut harvest, reducing manual sorting efforts by 95%. Classification reliability significantly exceeds manual sorting. The ability to automatically distinguish new quality classes of hazelnuts allowed Valeria to introduce new hazelnut-based products: premium nuts for gourmet cooking or uniformly sized nuts for chocolate bars.
3. New market potential: IFSYS showcased the machine at trade fairs, significantly advancing market entry into the agricultural sector.
4. Expandable design: At a sorting speed of 9 nuts per second, mechanics proved the limiting factor – not the AI. IFSYS thus created a promising foundation for potential serial production.

“I learned that clear categorization and classification clearly improve our market opportunities.” - Markus Helmerich, Valeria, Haselnuss Großbardorf GbR

Conclusion: A directly integrated, customized AI solution offers entirely new business opportunities.

Through the implementation of a customized, AI-based sorting solution, IFSYS opened entirely new market opportunities in the agricultural sector. The high efficiency and flexibility of the sorting machine provide a compelling alternative to manual sorting, enabling producers to work more efficiently and improve product quality.

“We want to enter serial production with this machine. Our customers can use it to sort not only hazelnuts but also beans, lupins, cashews, seeds, and more. The agricultural sector has a strong demand for such a sorting solution.” - Adelbert Demar, founder of IFSYS

Christian

Senior Software Developer

Arne Müller

General Manager

Would you like to learn more about integrating customized AI components into your solutions?

and discover how together we can revolutionize your production processes with artificial intelligence.