Detecting Diversified Agricultural Systems with High-Resolution Imaging and Deep Learning

• Grant number: 25/01750-0
• Support Opportunities: Scholarships in Brazil - Master
• Start date: April 01, 2025
• End date: March 31, 2027
• Field of knowledge: Physical Sciences and Mathematics - Geosciences - Physical Geography
• Principal Investigator: Édson Luis Bolfe
• Grantee: Victória Beatriz Soares Leandro
• 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

Brazilian agricultural production, characterized by extensive monoculture and intensive land use, has exacerbated social, economic, and environmental issues in recent decades. In this context, diversified agricultural systems have emerged as a sustainable alternative, integrating agricultural crops, woody tree species, and/or livestock in a planned manner. These systems offer benefits such as greater productive diversification, increased income generation, soil and water conservation, and strengthening the participation of local communities. The application of Deep Learning techniques, a subfield of Machine Learning, has expanded as a promising approach for the automated monitoring and mapping of diversified production areas. This project aims to develop a methodology based on remote sensing and Deep Learning algorithms for the identification and analysis of these areas. The study has as possible areas of action the Agrotechnological Districts (DATs) defined by the Center of Science for Development in Digital Agriculture (CCD-AD SemeAr Digital), covering Jacupiranga (SP), Breves (PA), Guia Lopes da Laguna (MS) and Boa Vista do Tupim (BA). The proposed methodology will be structured in three main stages: acquisition and processing of images from remote sensors to extract spatial information, collection and structuring of sample databases associating field data and georeferenced images, and statistical analysis and computational modeling based on Deep Learning to classify and monitor productive areas. This master's study seeks to contribute to the decision-making of small and medium-sized producers linked to the SemeAr Digital project, providing an automated tool for mapping and monitoring diversified agricultural systems. The results obtained may support public policy initiatives aimed at agricultural sustainability and technological innovation in the rural sector.