We use cookies on our website. Some are necessary for the operation of the website. You can also allow cookies for statistical purposes. You can adjust the data protection settings or agree to all cookies directly.
Data Protection Declaration
Imprint
Collaborative Project: AI-supported optimization of the application of NIR/MIR sensors in agriculture - subproject A (KIO-Sens)
Project
Project code: 28DK112A20
Contract period: 01.06.2021
- 31.05.2024
Budget: 329,692 Euro
Purpose of research: Applied research
Keywords: digital world, grasses, crop production, feed crop production, sensor technology, maize, AI Artificial Intelligence, animal nutrition
Within the framework of this project, two major problems which arise in agriculture due to the use of NIR/MIR sensors are to be addressed and tackled. This concerns 1) the AI based classification of spectra (e.g. for the automatic detection of sample types like maize vs. grass or for traceability) and 2) the AI based online sensor standardisation. Numerous factors influencing the spectra must be taken into account to improve the prediction accuracy, e.g. the correct selection of the substrate (e.g. corn vs. grass), the temperature etc. Currently, such factors are manually selected by the experienced user and as best as appropriate to the underlying model. In the project, the use of machine learning and AI algorithms is intended to improve the precision of prediction models, i.e. the AI-supported automatic weighting of the model influencing factors should be combined adaptively to achieve a more accurate prediction. The development and use of appropriate AI (machine learning) tools should be based on interpretable classification algorithms for categorical variables (e.g. learning vector quantizers, LVQ) and non-linear regression algorithms for numerical variables (e.g. Bayesian regression, lasso & ridge regression, quantile regression). At the same time, methods of so-called transfer learning will be investigated in order to realize a sensor-adapted, automatic standardisation of spectra and to make them accessible to a common AI prediction model. This development of generally applicable and cross-technology methods based on the findings of NIR spectroscopy can then be transferred to other areas of the agricultural process chain or the food industry.
Section overview
Subjects
- Crop Production
Framework programme
Funding programme
Excutive institution
LKS-Landwirtschaftliche Kommunikations- und Servicegesellschaft mbH
