Banca de DEFESA: RUDIARD NARDELLI

Uma banca de DEFESA de MESTRADO foi cadastrada pelo programa.
DISCENTE : RUDIARD NARDELLI
DATA : 13/12/2019
HORA: 09:00
LOCAL: IFC- Campus Araquari
TÍTULO:

APP FOR IDENTIFICATION OF THE MAIN COMMERCIAL SHRIMP SPECIES

PALAVRAS-CHAVES:

image processing; species identification; confusion matrix; shrimp.

PÁGINAS: 45
GRANDE ÁREA: Ciências Agrárias
ÁREA: Medicina Veterinária
RESUMO:

Despite the considerable commercialization of extracted and farmed shrimp, there are still no electronic devices for verifying shrimp species within the industry. Currently the recognition of the species within the industrial plant is made visually by a professional trained in quality control, as well as documents provided by vessels or producers. A possible solution to assist in this task would be the use of digital image processing techniques, one of which is the digital classification of images using neural networks, which identify and learn to recognize spectral patterns by analyzing the digital value of one pixel of the image. Confusion matrix construction is a technique used to analyze the performance of classifier systems. The objective of this work was to verify the potential of using neural networks to identify the main commercial shrimp species in Santa Catarina. Logistic regression was the method considered in the neural network model and stratified cross-validation for sampling. The model evaluation presented a value of 99.5% for the AUC (Area Under Curve) and precision of 93.7% when all the photos were analyzed together. In the test within the industry the evaluation of the device obtained average hit of 87.85%, with image bank using 3424 photos. The results showed that the app can be used as an auxiliary tool to classify commercial shrimp species.

MEMBROS DA BANCA:
Interno - 1760760 - CARLOS EDUARDO NOGUEIRA MARTINS
Externo à Instituição - GILBERTO CAETANO MANZONI
Presidente - 2356802 - ROBILSON ANTONIO WEBER
Interna - 1085177 - VANESSA PERIPOLLI