Desenvolvimento de metodologias para determinação do bem-estar e massa corporal de frangos de corte por meio de análise digital de imagens

• Autor(a) principal: Mogami, Cristina Akemi
• Data de Publicação: 2009
• Tipo de documento: Tese
• Idioma: por
• Título da fonte: LOCUS Repositório Institucional da UFV
• Texto Completo: http://locus.ufv.br/handle/123456789/656
• topic: Comportamento animal; Visão artificial; Animal behavior; Artificial vision; CNPQ::CIENCIAS AGRARIAS::ENGENHARIA AGRICOLA::CONSTRUCOES RURAIS E AMBIENCIA

Advances have been made in the field of image processing as a technology for development of a consistent response in relation to new standards in animal welfare. One of the advantages of artificial vision systems in relation to those of human vision is the ability to make exact measurements, based on pixel count. This system also has no influence on animal behavior, does not require a physical observer, and permits data collection over long time periods. Therefore, the overall objective of this work was to develop methodologies to evaluate welfare and body mass of broiler chickens by means of digital image analysis. Experimental analyses was carried out in two commercial poultry houses integrated with the Empresa Perdigão Agroindustrial S.A., in Rio Verde, Goiás, Brazil during the final growth stage of the birds. The two buildings were equipped with tunnel ventilation systems with nebulization, one applying water sprinkling over the roof. The black globe temperature index (ITGU) was related to the results obtained by the processing of digital images acquired during the experiment. The Scilab computer program was used for processing of the images and the Bayes Method was used for linearization. Results were presented in two articles: Article I Development of methodology for evaluation of boiler chicken welfare in commercial poultry houses by digital image analysis; Article II Development of methodology for determination of broiler chicken body mass gain in commercial poultry houses using digital image analysis. The ICRC proved to be a practical and non- invasive method, reducing the influence of the grouping effect of birds in the image, but was inefficient to identify properly feeding chickens. The ICEC allowed for greater reliability of image analysis results and revealed to have no direct relation between the ITGU and number of birds at the feeder in the CA and SA treatments. The algorithm for determination of the number of occupied partitions (ICEC) showed to be an objective parameter, resulting in an error percentage of 11.6% between manual classification and that of the algorithm. Nocturnal activity at the feeding troughs and water fountains was low between 8:00 pm and 9:00 pm, followed by a significant increase in the morning. Binary counting of pixels resultant of the digital images of the chickens grew throughout the studied period, showing the same behavior as the average chicken weight. The empirical model used to determine body mass of the chickens as a function of the number of pixels in the images showed to be adequate for broiler chicken between 21 and 42 days old, in the pixel range of 800 to 1400.