Advanced search
Start date

Deep learning for the detection of weeds in aerial imagery collected by UAVs

Grant number: 22/01259-7
Support Opportunities:Scholarships in Brazil - Scientific Initiation
Effective date (Start): May 01, 2022
Effective date (End): April 30, 2023
Field of knowledge:Interdisciplinary Subjects
Principal Investigator:Flávio José de Oliveira Morais
Grantee:Ronaldo de Oliveira Elias
Host Institution: Faculdade de Ciências e Engenharia. Universidade Estadual Paulista (UNESP). Campus de Tupã. Tupã , SP, Brazil


With the need to optimize agriculture aiming at productivity, producers have been increasingly willing to use new technologies in the field. Following this trend, this research offers alternatives for the mitigation of losses (and consequent increase in productivity) by exploring the problem of weeds in crops. More specifically, in traditional agriculture, among different measures to reduce the impact of weeds, pesticides are applied to control them. However, the exacerbated use of many pesticides can result in high production costs, as well as harm to human health and the environment. In this sense, this work considers new technologies of the so-called Smart Farms, such as Unmanned Aerial Vehicles (UAVs) and advanced techniques of Artificial Intelligence to detect weeds. In particular, the so-called Deep Neural Networks will be investigated for this application, which has been successfully employed in pattern recognition in images. Based on these architectures, we will discriminate weed patterns in aerial images collected by a UAV by segmenting the pixels. The studied and proposed methods/algorithms will be compared on real datasets and the results will be presented considering metrics, such as IoU (Intersection-Over-Union) and Dice Coefficient, which are typically adopted in the literature to evaluate the performance of these tools.(AU)

News published in Agência FAPESP Newsletter about the scholarship:
Articles published in other media outlets (0 total):
More itemsLess items

Please report errors in scientific publications list using this form.