Banca de QUALIFICAÇÃO: MARIA ANGÉLICA MARÇOLA

Uma banca de QUALIFICAÇÃO de DOUTORADO foi cadastrada pelo programa.
DISCENTE: MARIA ANGÉLICA MARÇOLA
DATA: 28/02/2025
HORA: 10:00
LOCAL: Google Meet: meet.google.com/zan-yijk-dzv
TÍTULO:

ASSESSMENT OF ASIAN SOYBEAN RUST SEVERITY THROUGH DIGITAL IMAGE ANALYSIS

PALAVRAS-CHAVES:

Phakopsora pachyrhizi; Computer vision; Machine Learning

PÁGINAS: 24
GRANDE ÁREA: Ciências Agrárias
ÁREA: Agronomia
SUBÁREA: Fitotecnia
ESPECIALIDADE: Melhoramento Vegetal
RESUMO:

The objective of this study was to develop an algorithm based on a customized computer vision model, utilizing open-source software, to quantify the severity of Asian soybean rust (ASR) in digital images of soybean leaflets. The proposed approach involves two processing stages. The first stage identifies healthy and diseased areas of the leaflet, while the second focuses on segmentation. Communication between these stages is facilitated by defining a hyperparameter based on the image texture, which guides the detection of symptom presence or absence. Validation was performed through manual annotation of symptoms and correlation with the model's outputs. To evaluate performance, 15 researchers assessed 50 leaflets each using conventional methodologies, and the agreement between these methodologies and the proposed model's results was calculated. Validation achieved a coefficient of determination (R²) of 0.99, indicating high precision in symptom distinction. Regarding severity assessments, both conventional methodologies demonstrated a tendency to overestimate symptoms. However, the use of standard area diagram proposed by Franceschi et al. (2020) improved the quality of symptom estimates. Without the use of standard area diagram, greater errors were identified, especially in the overestimation of symptoms present in the leaflets. Consequently, the proposed model demonstrates significant applicability for precise, rapid, and scalable identification of symptoms in soybean leaflets.

MEMBROS DA BANCA:
Interno - VINICIUS QUINTAO CARNEIRO (Membro)
Externo à Instituição - LORENA NUNES LACERDA - UGA (Membro)
Externo ao Programa - FERNANDA APARECIDA CASTRO PEREIRA - DAG/ESAL (Membro)
Presidente - ADRIANO TEODORO BRUZI (Membro)
Externo ao Programa - ADAO FELIPE DOS SANTOS - DAG/ESAL (Suplente)