The determination of product-relevant data is to be carried out using specific image processing. In addition to an RGB camera to evaluate the color information on the product surface, this also includes a spectral camera that enables a qualitative assessment, e.g. the detection of the fat content of turkey breast or the ripeness of potatoes. The resulting images as well as their resolution are evaluated by DIL and DFKI and corresponding parameters for the later transfer to the software system are defined. With the collected data, models are built to classify the goods with respect to their quality. The model will be developed to a point where it can be verified whether an improvement over previous methods can be reliably achieved. For other product types it is to be examined then, how well the models themselves, in addition, the infrastructure for providing the models can be transferred. During the later implementation and validation of the prototype, the software system for classification will be integrated into the overall structure by DFKI and the interfaces between classification algorithm and robotics will be established. Usable results of the project will be processed in a timely manner and demonstrated in practice-relevant magazines and events and via online media (2 demonstration videos will be produced). Here, the DFKI supports research partner DIL.
DIL Deutsches Institut für Lebensmitteltechnik e.V.
Sola Bonum GmbH
Wernsing Feinkost GmbH & Co KG
Heidemark Mästerkreis GmbH & Co. KG