Discente: Ellen Marques Giacometti

 

Data: 29 de maio de 2024

 

Hora: 14 h

 

Localização: Microsoft Teams

 

Link: https://teams.microsoft.com/l/meetup-join/19%3ameeting_OTA0ODYxZTQtYzVjYi00MGE4LWFjM2EtYmE2NDgwZjZkOWMz%40thread.v2/0?context=%7b%22Tid%22%3a%228eeca404-a47d-4555-a2d4-0f3619041c9c%22%2c%22Oid%22%3a%228d7ec457-a07f-4b62-9549-eac169ae67f6%22%7d

 

 

Resumo: The Tahiti lime accounts for a significant portion of Brazil’s fruit exports. The quality classification of fruits is a critical issue for agribusiness, as the price of fruit is determined by its quality. Autonomous fruit classification has become an essential process, as it allows for the rapid and efficient selection of fruits for export. This capability reduces costs and enhances competitiveness in the agribusiness market. In this context, the creation of an automatic classifier system for Tahiti lime images using computer vision has been proposed. This system is capable of classifying limes based on their basic characteristics, such as color, size, and possible deformations. To train the proposed system and evaluate the results of the applied machine learning techniques, a database containing hundreds of Tahiti lime images arranged in different ways under various lighting conditions was created. To expand the created database, data augmentation will be applied, where the initial sample quantity will be increased by using the original image and altering it to obtain another image, or by generating new samples through SMOTE. The results obtained with the proposed model meet the quality standards established by UNECE in terms of file size. The best classifier achieved satisfactory results in terms of maturity and presence of defects, achieving over 85% accuracy using Random Forest.

 

 

Presidente: Gabriel Matos Araujo (orientador, CEFET/RJ)

Diego Barreto Haddad (CEFET/RJ)

Amaro Azevedo de Lima (CEFET/RJ)