Use of artificial intelligence and deep learning in the food and agriculture domain

Abstract

The current thesis is an overview of the materials and the technics that are used in the Smart Farming sector. It focuses on the particularities of Smart Farming compared to other Artificial Intelligence applications, hence a basic knowledge of AI constitutes a prerequisite. At the first chapter, a definition and the basic idea of Smart Farming, as well as the necessary conditions for the emergence of this sector and the tasks that it is meant to cope, are represented. In continuation, a brief description of the majority of sensors and platforms (e.g., UAVs, tractors) used in the Smart Farming are analyzed, focusing on the advantages, disadvantages and particularities of each one. Then, a description of the communication technics amongst the nodes is provided, with an emphasis on the Low-power wide area network which is mostly used in agriculture. The thesis describes how software related tasks are tackled, like preprocessing technics and AI models. An object detection algorithm applied on orange fruits is implemented as an example of a Smart Farming application that could be used for autonomous fruit picking. Finally, a social and economic analysis of Smart Farming is provided, along with the obstacles it faces from being adopted by the small-scale farmers.