Imagine cameras that can see things the human eye cannot — the chemical makeup, moisture level, or hidden defects inside materials, all while products fly down a conveyor belt. MULTIPLE built affordable versions of these super-cameras that work across visible light and infrared, then connected them to AI that learns what "good" and "bad" look like. They proved it works on three very different production lines: rolling steel, making furniture, and producing chocolate. The result is a quality control system that spots problems before they become expensive waste.
Most manufacturers still rely on manual spot-checks or basic sensors for quality control, catching defects only after products are already made — leading to scrap, rework, and customer complaints. Hyperspectral cameras that can see both surface defects and chemical composition exist, but they have been too expensive and complex for routine factory use. Production managers need affordable, real-time monitoring that works on fast-moving lines without slowing anything down.
The project built cost-effective hyperspectral camera cores (VIS/SWIR, 0.4–1.7 µm), OLED-based spectrometers, and laser-based chemometric sensors covering up to 3.5 µm. These were paired with embedded deep learning models and an IoT cloud platform for orchestrating AI-based process optimization. All of this was demonstrated on 3 real production lines: a steel rolling mill, a furniture factory, and a chocolate production line.
If you are a steel producer dealing with surface defects or inconsistent material quality on your rolling lines — this project developed a hyperspectral monitoring system demonstrated directly on a rolling mill production line. It uses AI-driven spectral analysis to detect defects in real time, reducing scrap and rework before bad product reaches your customers.
If you are a furniture manufacturer struggling with inconsistent wood quality, color mismatches, or hidden material defects — this project built and demonstrated a multimodal spectral monitoring system on a furniture manufacturing line. It classifies wood properties across the visible and shortwave infrared range (0.4–1.7 µm) to sort and grade materials automatically.
If you are a chocolate or food producer facing batch-to-batch inconsistency or contamination risks — this project demonstrated a cost-effective spectral sensor system on a chocolate production line. It uses laser-based chemometric analysis up to 3.5 µm wavelength to monitor composition in real time, catching quality deviations before they ruin an entire batch.
The project's core goal was to develop 'cost-effective' sensor solutions — specifically cheaper hyperspectral cameras, OLED-based spectrometers, and compact dual-aperture imagers designed for volume production. Exact pricing is not published in the project data, but the emphasis on cost-effectiveness and volume-ready camera cores suggests pricing well below current laboratory-grade hyperspectral systems. Contact the consortium for specific quotes.
Yes. The system was demonstrated on 3 real production lines — steel rolling, furniture manufacturing, and chocolate production — not in a lab. It includes embedded deep learning models for real-time regression, classification, and control, plus IoT-native cloud architecture designed for continuous industrial operation.
The consortium of 20 partners across 9 countries jointly developed the technology. With 17 industry partners (85% industry ratio) and 8 SMEs involved, several consortium members are likely positioned to commercialize or license specific components. Contact the coordinator (ASOCIACION DE INVESTIGACION METALURGICA DEL NOROESTE in Spain) for IP and licensing discussions.
The sensors cover visible light through shortwave infrared (0.4–1.7 µm) and mid-wave infrared up to 3.5 µm. This broad range means you can detect surface defects visually AND analyze chemical composition beneath the surface — something standard cameras or simple sensors cannot do.
The project ran from December 2019 to April 2023 and completed all 3 industrial demonstrations. As an Innovation Action with finished demonstrators, the technology is at pilot stage. Some components — particularly the camera cores designed for volume production — may already be commercially available from consortium partners.
The system was designed as IoT-native with cloud connectivity and edge computing, meaning it can be integrated alongside existing equipment rather than replacing it. The modular approach — separate sensor hardware, embedded AI, and cloud orchestration — allows you to add monitoring capability to specific points on your line without a full rebuild.
The chocolate production line demonstration shows the system was designed with food manufacturing in mind. Based on available project data, the laser-based spectral sensors operate without contact with the product. Specific regulatory certifications would need to be confirmed with the consortium partners who built the food demonstrator.
This is a heavily industry-driven consortium: 17 out of 20 partners are from industry, giving it an 85% industry ratio — unusually high even for Innovation Actions. With 8 SMEs and partners across 9 countries (Belgium, Switzerland, Germany, Greece, Spain, France, Italy, Poland, Portugal), the project combines sensor manufacturers, AI developers, and end-user factories. The coordinator is a Spanish metallurgical research association, which explains the steel use case. The absence of universities is notable — this was built to deploy, not to publish papers. For a business looking to adopt this technology, the high industry ratio means multiple potential suppliers and integrators already exist within the consortium.
The coordinator is ASOCIACION DE INVESTIGACION METALURGICA DEL NOROESTE in Spain — a metallurgical research association. SciTransfer can identify the right contact person and facilitate an introduction.
Want to know if MULTIPLE's spectral monitoring technology fits your production line? SciTransfer can connect you with the right consortium partner for your specific use case — steel, wood, food, or beyond.
