Analysis of stresses in agricultural crops using deep learning and remote and proximal sensing imagery

• Grant number: 25/05985-2
• Support Opportunities: Scholarships in Brazil - Post-Doctoral
• Start date: June 01, 2025
• End date: May 31, 2027
• Field of knowledge: Interdisciplinary Subjects
• Principal Investigator: Jayme Garcia Arnal Barbedo
• Grantee: Letícia Ferrari Castanheiro
• Host Institution: Embrapa Agricultura Digital. Empresa Brasileira de Pesquisa Agropecuária (EMBRAPA). Ministério da Agricultura, Pecuária e Abastecimento (Brasil). Campinas , SP, Brazil
• Associated research grant: 22/09319-9 - Center of Science for Development in Digital Agriculture - CCD-AD/SemeAr, AP.CCD

Abstract

The rapid diagnosis of stresses in economically valuable plants is essential to ensure food security and prevent greater losses resulting from the spread of such conditions. Two main challenges can hinder this goal: (1) the continuous monitoring of all plants by individuals capable of detecting stress is, in most cases, unfeasible; and (2) in many cases, the person who identifies the symptoms lacks the necessary knowledge to determine their cause.Although some solutions have explored technology as a facilitator for rapid diagnosis-usually involving a reference database that users can consult-the emergence of truly automatic systems has been quite slow, especially considering the importance of the issue. This is not due to a lack of interest from researchers, but rather the absence of a sufficiently comprehensive dataset that would enable the development of truly robust diagnostic methods. Therefore, building datasets that allow for the full validation of proposed methods is just as important as developing the techniques themselves.Thus, the present project has three main objectives:1) Expand existing datasets with images covering a wide range of conditions, so that the developed models are sufficiently robust to the diversity of real-world scenarios. These images will be captured using RGB cameras embedded in mobile phones and unmanned aerial vehicles (UAVs).2) Develop methods capable of providing a reliable diagnosis based on images captured using the specified sensors.3) Build image databases using other types of sensors. In particular, the project aims to build a hyperspectral image database, as this technology holds significant potential for detecting stress before it becomes visually apparent.The project will focus on areas selected as Agrotechnological Districts (DATs) under the Semear Digital project (Center for Science for Development in Digital Agriculture - CCD-AD/SemeAr, FAPESP process no. 2022/09319-9).