On the impact of the stem electrical impedance in neural network algorithms for plant monitoring applications

Smart agriculture offers an environmental-friendly path with respect to unsustainable farming that depletes the nutrients in the soil leading to a persistent degradation of ecosystems caused by population growth. Artificial Intelligence (AI) can help mitigate this issue by predicting plant health status to reduce the use of chemicals and optimize water usage. This paper proposes a custom framework to train neural networks and a comparison among different models to point out the impact and the importance of the stem electrical impedance in addition to environmental parameters for plant monitoring applications. In particular, the paper demonstrates how stem electrical impedance improves the accuracy of the proposed neural network application for plant status classification. The data set is composed of electrical impedance spectra and environmental data acquired on four tobacco plants for a two-month-long experiment. In this paper, we describe the acquisition system architecture, the feature composition of the data set, a general overview of the developed framework, and the training of the neural networks showing the different results considering both the stem impedance and the environmental parameters.