Vision-based system for detecting grapevine yellow diseases using artificial intelligence

Y. Ampatzidis, A. Cruz, R. Pierro, A. Materazzi, A. Panattoni, L. De Bellis, A. Luvisi
Grapevine yellows (GY) of grapes, a critical threat to grapevines because of the severe symptoms and the lack of healing treatments, has been detected worldwide. The detection of GY diseases is a very difficult and time consuming task, and relies on symptoms identification, which are very similar with other diseases. Additionally, the analysis of asymptomatic GY-infected grapes could lead to high rates of false-negative due of the low concentration of the pathogen in the host. Herein, we present a supporting vision-based tool for GY disease detection using artificial intelligence (AI) and machine learning (ML). Leaves of bois noir-infected plants (previously tested by qPCR) were collected in July-October, 2017. Grapevine yellow was detected in a data set of 322 images and six diseases. Other than grapevine yellow, the diseases include downy mildew, esca disease, grapevine leafroll, powdery mildew and Stictocephala bisonia. A linear support vector machine (SVM) classified features from a pre-trained convolutional neural network - AlexNet trained on ImageNet. The system obtains a 95.23% accuracy and a Matthew's correlation coefficient of 0.832. For reference, a baseline system with local binary patterns (LBP) and color histogram with a SVM obtains only 26.7% and -0.124, respectively. Our work shows promise for automatic detection of grapevine yellow by computers. Future work will focus on improving the sensitivity of the system and implementation on drones with Nvidia Jetson. This system could reduce the rate of false positive/negative in large-scale vineyard monitoring.
Ampatzidis, Y., Cruz, A., Pierro, R., Materazzi, A., Panattoni, A., De Bellis, L. and Luvisi, A. (2020). Vision-based system for detecting grapevine yellow diseases using artificial intelligence. Acta Hortic. 1279, 225-230
DOI: 10.17660/ActaHortic.2020.1279.33
grapevine yellows, artificial intelligence, machine learning, deep learning, vision-based, disease detection

Acta Horticulturae