Machine Learning in agrifood

Machine Learning in agrifood

From monitoring food product quality to worldwide food safety issues: there is an overwhelming amount of data available that could support agrifood companies to make better-informed decisions. But how do you find precisely the right information when data volumes are so large? Wageningen Food & Biobased Research helps companies extract and analyse data for use in decision support.

Artificial intelligence

Can we automatically map specific food products to general nutrient data values? What does the number of food safety incidents in the world say about emerging issues for an individual food company? Is it possible to predict fruit quality based on cultivation data? An increasing amount of data is available that provides opportunities for smart decision-support systems. Advanced artificial intelligence (AI) or, more specifically, machine learning, plays a crucial role in unlocking and analysing these data. A well-designed AI system can automatically recognise specific patterns in the data that are key to, for example, evaluation of food safety risks or quality prediction, while being unaffected by data noise.

Deep learning and Bayesian networks

Wageningen Food & Biobased Research provides artificial intelligence solutions that help companies to analyse data and create new algorithms for smart software. We are experts in combining methods and tools from machine learning, including statistical learning, text data mining, and deep learning.

In agrifood, existing data is often not immediately ready for use and specific product and process expertise is needed to organize, interpret and analyse them correctly. We provide in-house developed semantics technologies that help companies organize and solve such issues. We combine our knowledge of machine learning with knowledge modelling methods and a broad expertise in agrifood, covering the whole production chain. This unique background allows us to provide innovative solutions that can be incorporated in web applications, smartphone apps and embedded systems. For example, we apply Bayesian Belief Networks to combine expert knowledge and semantically enriched datasets to create smart algorithms.

Smart chains, smart search

Quality prediction of fresh fruits and vegetables that allows suppliers to make well-considered decisions about storage, transport and markets. This is the outcome of GreenCHAINge (2016-2019), a project that aims to create a smart supply chain. We apply machine learning to predict product quality.

Farmers can efficiently identify innovations in agriculture and forestry with the smart search engine we developed within VALERIE (2014-2017), a project with over ten different European partners. Information on new farming techniques is usually difficult to find on the internet. The work will be continued in FAIRshared, which focuses on digital innovations.

The development of cash registers that recognize products by their shape and colour is now possible thanks to a machine-learning solution we developed for the Swedish company ITAB (2008-2019). The system automatically recognizes products going through a supermarket checkout, with 99% accuracy, using different pattern-recognition techniques.