Is it possible to automate quality control of rapeseed with a mobile app? This was the question the German Institute of Food Technologies (DIL Technologie) chose to investigate in its S3FOOD project.
Prior to the project, rapeseed quality control is carried out by employees, who evaluate, for example, dehulling quality and the degree of hull separation in processing. This manual process is both expensive and fails to ensure a standardised outcome, due to the subjective perceptions of each member of staff.
The institute is collaborating with VT-Engineering to investigate how the integration of AI in the classification of image data can support a high-performing app for automated quality control. The aim is to demonstrate the general technical feasibility of an app with sufficient classification accuracy for two to three product classes.
Final report summary
In the project, a mobile APP for the classification of rapeseed was successfully implemented and tested. By integrating a neural network for classification, a high accuracy and robustness against environmental influences, such as flash light, shadows, etc., could be achieved. A preliminary version of the mobile app using classical image processing algorithms could not show such robustness, so the need for neural networks for classification was also shown by this project.
The next steps would be to test the system in an industrial environment and to integrate it into an overall system. For this, the system needs to be optimised, adapted and extended to enable operation over a conveyor system, the direct implementation of control commands in the overall system and real-time use.
The project has enabled DIL Technologie GmbH to work with the creation of a mobile app and the implementation of neural networks. This offers the company new opportunities in this field. Furthermore, the interesting market of rapeseed processing could be explored.