Determination of Machine Learning and IOT Enabled Complete Architecture for Agriculture Applications

Abstract

Tradition agriculture models have certain limitations, such as the risk cost, actual implementation without required quality information. For any crop and its productivity improvement needs the soil supplements such as moisture, nitrogen (N), phosphorus (P), potassium (K). In order to overcome such limitations of traditional approach we develop a model to predict the soil fertility level and productivity forecasting using deep learning and remote sensing. Current advancements in technology that offers foretelling form, which improves the state of the art precision agriculture. The use of machine learning (ML) techniques with IoT devices in various fields of discipline has increased, in the last decade. The mounting accessibility of soil data that accelerate by adoption of IoT enabled ML techniques to analyze and improved productivity.  In the first step, we are establishing the physical environment, where IoT devices are placed in the fields to capture the soil properties. Second step, we develop Back-propagation neural network, an algorithm of machine learning and deep learning, model dynamic way to predict soil properties and evaluate from raw soil field the input data receive from first phase. Currently available Internet of Things (IoT) devices and connectors for wireless communication with sensors are applicable for different purpose of agricultural field work such as preparation of soil, water management, and crop growth status. IoT and deep learning based state-of the-art agriculture architecture identify the traditional limitation and give the appropriate solution. The main study objective of this research paper is to look at the use of deep learning, particularly back propagation neural networks, which use inputs from various IoT devices to predict soil properties from spectral (raw) data from organic soils.