Kamerabaseret bestemmelse af sygdomstryk sukkerroer med kunstig intelligens

Conclusion

Based on three years of results achieved in this project:
• It is possible to collect colour images of sugar beet using an RGB camera mounted on a field robot in sufficient quality for a human to recognize the common diseases rust, powdery mildew, and Cercospora leaf spot.
• After annotating the images, it is possible to train and validate a deep neural network to recognize sugar beet leaves, background, and diseases (rust and partly powdery mildew; Cercospora leaf spot was not present in test data).
• It is possible to monitor the development of the disease in sugar beet over time by analyzing the collected images using the trained network.
• We can identify areas in the field experiments where diseases start and develop from.