Machine learning-based multi-modal data fusion and growth modeling for soybean production improvement

• Grant number: 24/15430-5
• Support Opportunities: Research Program on Global Climate Change - Thematic Grants
• Start date: January 01, 2025
• End date: December 31, 2027
• Field of knowledge: Physical Sciences and Mathematics - Computer Science - Computing Methodologies and Techniques
• Principal Investigator: Zhao Liang
• Grantee: Zhao Liang
• Principal researcher abroad: Huaqiang Yuan
• Institution abroad: Dongguan University of Technology, China
• Host Institution: Faculdade de Filosofia, Ciências e Letras de Ribeirão Preto (FFCLRP). Universidade de São Paulo (USP). Ribeirão Preto , SP, Brazil
• Pesquisadores principais: José Baldin Pinheiro ; Renato Tinós
• Associated researchers: Alexandre Cláudio Botazzo Delbem ; Amaury Burlamaqui Bendahan ; Antonio Mauro Saraiva ; Bruno Sérgio Vieira ; Ednaldo José Ferreira ; George Deroco Martins ; Jayme Garcia Arnal Barbedo ; Joaquim Cezar Felipe ; Julio Cezar Franchini dos Santos ; Juscelino Izidoro de Oliveira Júnior ; Lucas Prado Osco ; Lucio Andre de Castro Jorge ; Luiz Otavio Murta Junior ; Mirela Teixeira Cazzolato ; Murillo Guimarães Carneiro ; Rafael Moreira Soares ; Ricardo Marcondes Marcacini ; Rodrigo Yoiti Tsukahara
• Associated scholarship(s): 25/11161-2 - High throughput phenotyping and machine learning for selection of soybean genotypes resistant to stink bug complex in different maturity groups, BP.PD; 25/08971-2 - Development of a WEB Decision Making System for Soybean crops by Artificial Intelligence Techniques using Multi-spectral and Hyper-spectral Images and LIDAR cloud points using UAVs and Satellites images, BP.TT

Abstract

Accurate production management decisions for soybeans are crucial for increasing yield and improving harvest quality. However, the complex canopy structure of soybean plants, characterized by mutual shading, complicates the accurate acquisition of phenotypic data. Additionally, soybean growth is intricately linked to environmental factors, which are increasingly variable due to climate change. The diversity of production monitoring data, coupled with varied and complex scenarios across different regions, poses significant challenges for modeling soybean production management and decision-making. To address these challenges, this project leverages a collaborative effort between Chinese and Brazilian research teams, focuses on utilizing multi-modal data gathered during the soybean production process and employs 3D reconstruction, phenotypic analysis, multi-layer complex network modeling, and domain-adaptive learning as its theoretical and technical foundation. The research aims to develop key technologies that leverage machine learning and multi-modal data fusion to enhance soybean yield and improve harvest quality. The main tasks include: 1) Developing methods for high-quality 3D reconstruction of soybean plants using sparse multi-view images to achieve precise structural modeling; 2) Advancing phenotyping technology for soybean plants based on multi-modal data such as images, point clouds, and spectral information to enhance phenotype accuracy; 3) Creating an intelligent soybean management decision model that integrates multi-modal data fusion to optimize production management; and 4) Designing domain-adaptive models for yield prediction and plant analysis to improve the generalization of management strategies across diverse environments. By integrating machine learning with multi-modal data fusion, this this Brazil-China collaborative project aims to elevate soybean production management and provide valuable insights into the application of these technologies. (AU)