Identification of pests and diseases in agriculture - conventional models with artificial intelligence

Abstract

Crop production in agriculture has an important impact on the global economy. In recent years, effective production has played a key role, considering that the world population is continuously growing, increasing the risk of malnutrition. Plant diseases are one of the biggest threats, being often responsible for crop losses. Traditional plant and pest disease identification models have been developed, which rely on computer vision algorithms and weather data. On the other hand, Artificial Intelligence has gained popularity inrecent years, also due to better hardware capabilities, offering new ways to improve these models.This thesis aims at comparing conventional plant disease identification models with Artificial Intelligence models.First, traditional plant disease identification models are identified and explored. Next,Artificial Intelligence models are defined. These models use the same data as the traditional ones, making a comparison possible. In order to evaluate the models, a tool is implemented. It allows applying different models to identify and monitor plant diseases.It is built with extensibility in mind, allowing implementing and integrating new models.Finally, conventional and artificial intelligence models are compared. It is shown how Artificial Intelligence can improve plant disease identification and what benefits it brings to it, not only in terms of disease identification accuracy, but also allowing for an automated solution that can be used in practice.