Introduction

Agriculture has been the sole reason for the transformation of nomadic life to the settled life of humans. With passing times, it not only became the source of the main food supply for the ever-growing population but also acted as raw material feeder industry for most of the industries, leading to the overall development of humankind. While the industries grew and accommodated new technologies, the farmer himself had no or limited access to data on climate, water supply, energy availability, demands in the market, movements of the pests like locusts or even bigger animals and plant diseases. This all has led to the fact that even if farming as an industry is the source of livelihood it has lagged miles behind in terms of integration with data analytics and other relevant techniques, affecting the overall yield.

Smart agriculture encompasses the use of many on-field sensors like cameras, soil moisture sensors, temperature sensors, pH sensors, gas and smoke sensors and light intensity sensors. The data from the sensors can be collected either through a direct connection to the computer or can be saved to the cloud. If the data are saved to the cloud, it may be available as a reference as well. The data related to the market can be collected from various sources available on the internet.

The data collected from various sources can now be put through rigorous analysis. This is where machine learning comes into actual play. The data can be standardized and finally put through various learning algorithms for clustering, classification, prediction and finally control and monitoring of many variables. The control here implies controlling various relays, motors and other actuators for manipulating the immediate environment. This is aimed to help the farmer in proper maintenance of the crops and the livestock with a lesser amount of personal intervention. The data and various interpretations thus generated can be used to help other farmers as well in real time. Technological advancements have proved themselves to be highly reliable in increasing the yield. If technological advancements are aided with accurate analysis of data, then agriculture like any other sector can be organized and made profitable.

The related state-of-the-art technologies to date. Then a case study is proposed for detecting various diseases in tomato leaf. Several machine learning algorithms were considered for this purpose. Deep learning algorithm is found to be one of the effective techniques though different approaches have been used. Also, neural network plays a vital role to perform the task.