IDENTIFICAÇÃO E CLASSIFICAÇÃO DE MASTITE EM VACAS LEITEIRAS ATRAVÉS DA ESPECTRORADIOMETRIA

Abstract

The milk production represents a sector of great importance for Brazilian agribusiness, both for financial issues linked to the activity and cultural aspects related to producers. Despite the large volume produced, there are several difficulties faced by livestock farmers, one of the main ones being mastitis, inflation caused in the mammary glands of cows. Therefore, attention should be paid to new methods that help identify milk contaminated with the disease, in a practical and reliable way. The work addresses the application of spectroradiometry in the identification and classification of mastitis in dairy cows, comparing it with the traditional California Mastitis Test (CMT). To carry out the analyses, 68 milk samples from Girolando cows were used, collected on a rural property located in the municipality of Costa Rica/MS. The samples were analyzed for CMT and were then subjected to spectroradiometry. In the spectral analysis of the samples, two sensor readings were carried out on each sample, in which the spectral information was obtained using a spectroradiometer (FieldSpec 3 Jr da Analytical Spectral Devices, Longmont, CO, EUA). The values obtained were intended for machine learning, including algorithms such as artificial neural networks (ANN), J48 decision tree (J48), REPTree (DT), random forest (RF) and support vector machine (SVM). A logistic regression (RL) analysis was used as a control model. The results showed that spectroradiometry was effective in identifying mastitis, with greater accuracy in the visible (B-VIS) and near-infrared (C-NIR) spectral bands. However, the technique presented limitations in identifying subclinical mastitis and in samples without significant infection, in addition to presenting difficulties in differentiating levels of subclinical mastitis. Among the algorithms tested, SL demonstrated greater accuracy. It is concluded that spectroradiometry has potential as a complementary tool for diagnosing mastitis.