Detecting Severe Plant Water Stress with Machine Learning in IoT-Enabled Chamber

Abstract

Smart farming is revolutionizing the agricultural sector through the application of Information and Communication Technologies (ICT), aiming to enhance efficiency and ensure sustainability. This paradigm shift represents a pivotal advancement from traditional farming practices, setting a new precedent for the future of agriculture. By leveraging the Internet of Things (IoT), Artificial Intelligence (AI), and robotics, smart farming enables continuous and dynamic assessment of both environmental conditions and intrinsic plant responses. The extensive datasets collected can then be analyzed using machine learning (ML) to identify features that are essential indicators of specific plant status. In this work, we present a comparative analysis of different ML models designed for the detection of severe water stress conditions in plants. To monitor plant status in real-time, electric signals reflecting the bioelectrical activity of plants are measured using the Vivent biosignal amplifier. The bioelectrical activity is altered under water stress conditions, enabling the detection of this stimulus. Additionally, we engineer a controlled growth chamber with automated monitoring and management of critical growth factors including lighting, temperature, and irrigation. The results reveal that the eXtreme Gradient Boosting (XGB) classifier demonstrates superior performance, with an accuracy of 96% in distinguishing between healthy and water-stressed plants under controlled conditions, indicating its potential as an asset for precision agriculture applications.