APPLICATION OF MACHINE LEARNING ALGORITHMS AND REMOTE SENSING IN PREDICTING PEANUT HARVEST LOSS

• Grant number: 23/12653-0
• Support Opportunities: Scholarships in Brazil - Scientific Initiation
• Start date: February 01, 2024
• End date: January 31, 2025
• Field of knowledge: Physical Sciences and Mathematics - Geosciences
• Principal Investigator: Rouverson Pereira da Silva
• Grantee: Gabriel Pereira Costa
• Host Institution: Faculdade de Ciências Agrárias e Veterinárias (FCAV). Universidade Estadual Paulista (UNESP). Campus de Jaboticabal. Jaboticabal , SP, Brazil
• Associated research grant: 21/06029-7 - High resolution remote sensing for digital agriculture, AP.TEM

Abstract

Peanuts are an agricultural crop of great relevance in Brazil, both for the economy and for human and animal nutrition. However, the mechanized peanut picking operation often faces challenges related to losses in the field, including visible losses, such as pods lost during harvest and deposited on the ground, and invisible losses, which represent pods that remain under the ground. These losses represent a serious economic problem for farmers and industry. The objective of this project is to evaluate the performance of two machine learning algorithms, Multiple Linear Regression and Random Forest, in predicting visible and invisible losses during the mechanized uprooting operation of the peanut crop. The first step is to create a comprehensive dataset that will contain essential information such as soil texture, topographic indices and vegetation index. This data will serve as input for machine learning algorithms. Carefully collecting and processing this information will ensure the quality of the training set. Next, an algorithm will be developed with the ability to predict peanut harvest losses using soil characteristics and maturation index, such as NDVI, EVI, MNLI and GNDVI that are estimated by aerial remote sensing. These indices provide crucial information about vegetation health and land cover, being useful for monitoring large areas efficiently. Multiple Linear Regression and Random Forest models will be implemented and adjusted to make the most of information about soil texture and topographic indices. This project aims to provide farmers with effective tools to manage peanut harvest losses, contributing to improving productivity and reducing economic losses. Furthermore, the incorporation of remote sensing techniques will also be explored to enrich the dataset and further improve the accuracy of predictions.