Banca de DEFESA: ALINE SILVA RAMOS

Uma banca de DEFESA de MESTRADO foi cadastrada pelo programa.
DISCENTE : ALINE SILVA RAMOS
DATA : 31/05/2019
HORA: 08:30
LOCAL: Sala de aula do 7º andar da Escola Politécnica
TÍTULO:

SOMATIC CELL COUNT IN BUFFALOS MILK USING A FUZZY CLASSIFIER AND IMAGE PROCESSING TECHNIQUES

PALAVRAS-CHAVES:

SOMATIC CELL; BUFFALOS MILK; FUZZY; IMAGE PROCESSING TECHNIQUES.

PÁGINAS: 99
GRANDE ÁREA: Engenharias
ÁREA: Engenharia de Produção
RESUMO:

Milk production of buffaloes and their derivatives has been increasing in Brazil and in the world, together with the increasing demands on its quality standard. Mastitis, inflammatory disease of the mammary gland (GM), is responsible for qualitative and quantitative losses in relation to the milk produced. The somatic cell count (CCS) in milk is the main biomarker for both detection and evaluation of milk quality. CCS is traditionally determined by laborious methods consisting of the visual observation of cells in milk smears through the microscope. This traditional technique is exhaustive and has an inherent degree of subjectivity in that it is subject to human interpretation in relation to the identification and recognition of cells. For this reason, this research proposes and presents an automatic method for counting somatic cells in buffalo milk which includes, among others, the application of a Fuzzy clustering method and image processing techniques. Unlike other similar works, the Fuzzy C-Means classifier was used in the preprocessing stage of the images and not in the segmentation stage of the images. This approach enabled the separation of the somatic cell images (objects) present in buffalo milk in clusters that showed similarities in relation to the color intensity, allowing a better posterior application of processing techniques such as thresholding, segmentation and image recognition (interpretation of somatic cells). Three methods of thresholding were evaluated and compared, and the Watershed Transform was used to separate cells closely together, which contributed to the correct identification and counting of the same. Finally, a comparison was made between the results obtained by manual counting by the direct microscopic technique and by the method proposed in this work. A non-parametric statistical test (Kruskal Wallis) was used, which proved to obtain consistent counts results. The use of a Fuzzy classifier in the preprocessing of the images was a potential and efficient alternative for the classification of images in clusters that show similarity in color intensity. which provides a better performance of the thresholding process and consequently the somatic cell count in the images.

MEMBROS DA BANCA:
Presidente - 2199115 - CRISTIANO HORA DE OLIVEIRA FONTES
Externo ao Programa - 1867405 - KARINA MEDICI MADUREIRA
Externo ao Programa - 2130353 - RICARDO ARAUJO RIOS
Externo à Instituição - VIVIANI GOMES