Development of automatic methods for evaluating dairy cows behavior in confinement system

Author(s)

Silvia Regina Lucas de Souza

• Document type: Doctoral Thesis
• Press: Campinas, SP.
• Institution: Universidade Estadual de Campinas (UNICAMP). Faculdade de Engenharia Agrícola
• Defense date: 2006-12-19

Examining board members

Irenilza de Alencar Nääs; Evaldo Antonio Lencioni Titto; Irineu Arcaro Junior; Soraia Vanessa Matarazzo; Antonio Mauro Saraiva

• Advisor: Daniella Jorge de Moura; Irenilza de Alencar Nääs

Abstract

The current tendency of the milk production sector is to become more precise and economically feasible. It is known that differences in microclimate cause changes in animal behavior; however, conventional measures taken nowadays are evasive and can produce misinterpretation of the results. The current experiment was carried out in a commercial farm in São Pedro- SP, where milking cows were monitored using a total of six color micro-cameras with spatial resolution of 420 horizontal lines and 120 degrees lenses for behavior analysis. Indirect visual observations were performed parallel to the software analysis, as well as instantaneous indirect visual observations. The barn was divided into 18 quadrants in order to perform the climatic distribution mapping. The data gathered were processed using the software Surfer ® 6.0 to check for homogeneity. The primarily objective of this work was the development a milking cow?s behavior evaluation methodology as a function of freestall housing micro-climate, using image analysis. The results obtained through the multivariate analysis of the data from instantaneous observations pointed through the graphic of main components, the behaviors considered important for obtaining visually expressive responses when affected by the housing environment, such as: drinking water, feeding and laying down. The climatic distributions within the barn were not homogeneous when the cooling systems were not on. Correlations with the preset thermal values presented by the animals through the main component charts and THI from indirect visual instantaneous observations were performed and was concluded that the methodology for computational vision has been developed with few limitations and pending adjustments on its algorithms (AU)