ABSTRACT The higher productivity from soil can be achieved through Wireless Sensor Networks (WSNs) technology; for this reason, the implementation of WSNs in precision agriculture is increasing day by day. Among the different technologies for crop monitoring, WSNs are recognized as a powerful option for collecting and processing data in the agricultural domain, with low cost and low energy consumption. The ultimate aim of this paper is to predict the precise values of soil nutrients nitrogen, phosphorus, and potassium (NPK) using a machine learning (ML)-based mathematical model formulation. The proposed system will use Global System for Mobile Communications (GSM) technology to automatically monitor soil parameters. It transfers soil data over the mobile network using a GSM modem. In the hardware part, sensor node units were developed to fetch and load data through the energy-efficient GSM, utilizing a bi-directional multi-level text messaging option to improve the alarm system’s efficiency. The system contains sensors such as potential of hydrogen and humidity (PH), which are used to monitor soil data collected from the sensors by the microcontroller. Additionally, as a novel contribution, machine learning based on regression models is applied to the NPK data. A novel data interpolation approach is proposed for data oversampling. The R2 values are compared, achieving nearly 99.5% accuracy in all cases. This paper focuses on various challenges and scalability issues encountered in research and discusses different results obtained by monitoring various soil parameters in a real-time agricultural environment.
Read full abstract