Together with the Fraunhofer Institute, we are working on the development of image recognition systems that automate the hardwood identification in paper by analysing microscopic images. The first step was to develop a methodology for the systematic generation of a large dataset.
The creation of the first data set consisting nine hardwood genera and the training of different AI models with these data have now been published. Model number one detects the cell type that is crucial for identifying hardwoods. Model number two classifies the cells - the actual identification of the wood genera. Various models were used for validation and the algorithms were optimised. Visualised in a confusion matrix, the results show that the genera, which can be clearly distinguished by people trained in wood anatomy, are also classified mainly correct by the AI.
In future, this should make it easier to control global timber flows to protect forests. The research project is funded by Fachagentur Nachwachsende Rohstoffe e.V. (FNR) and Federal Ministry of Food und Agriculture (BMEL).
Contact:
Dr. Andrea Olbrich
Dr. Stephanie Helmling
Jördis Sieburg-Rockel
Stephanie Wrage
More informations:
To the DOI link of the publication
![[Translate to English:] Logo des Bundesministerium für Ernährung und Landwirtschaft](/media/allgemein/logos/BMEL_Logo.svg){kind=logo}