HyperCook: AI multispectral imaging for smart online monitoring of baking process

Challenge & Solution

The baking industry is one of Europe’s earliest industries however modern baking industry faces challenges of achieving increasing efficiency, reducing waste while ensuring a high quality product is delivered in a timely, consistent matter to their customers.

HyperCook aims to bring the insights provided by exploiting CSEM’s Hyperspectal camera along with their domain expertise in Hyperspectral imaging systems to the industrial baking process to gain additional insights into the cooking process and enable smarter industrial baking ovens to consistently produce the perfect high quality product.

By combining multispectral imaging with relevant and robust algorithms, Hypercook will be able to monitor the biscuit baking process in a static or moving scene configuration. The solution may be later fully integrated and offer communication with the oven assuring a complete autonomous functioning as part of an integrated bakery management system.

A main challenge to address is the products variability, even within one type as biscuits (relief, shape, impact of the recipe, marking, etc), and also user-experience driven quality experience (visual, taste, texture). Well adapted data treatment and machine learning tools developed by GreenTropism are key to address these challenges and provide a data driven decision-making system.

Solution

Greentropism aim to develop a prototype that combines hardware, software and hyperspectral imaging components along with their domain knowledge expertise, to provide an integrated system to monitor and control the cooking process. Such a system may be installed inline in the baking process and enable real-time feedback control of the cooking process. The system will provide information at different levels : from the user perspective, it will automate the determination of cooking stage by applying embedded machine learning algorithms ; later on, for an even smarter integration, it can be further interconnected with industrial ovens, in order to adapt the protocol (time, temperature, humidity, speed of the conveyor belt etc.) as part of the perfect recipe.

FED4SAE Support

Greentropism is specialized in spectroscopy data analysis and machine learning’s which neatly compliments CSEM’s capacities and expertise in multispectral imaging and product integration. Intel provides the Compute platforms (IPC) to enable the secure integration of the various components as part of an integrated baking or manufacturing process. Insights in innovation management were contributed by Blumorpho which is important for product development as well as to consider the right business models and market approaches.

Impact

Machine learning applied to multispectral camera offers a wide range of potential applications in food & beverage as well as in health, cosmetics, agriculture etc. For example, after developing its expertise on biscuit baking levels, Greentropism will be able to extend applications to pastry, rotisserie, precooked frozen food processing etc. Developing those CPS is beneficial to the market as it allows a more dynamic, cost reducing and homogeneous production. Hypercook is leading the way in ensuring quality and consistently of our food while delivering a greener industry with reduced waste and energy consumption.