Monitoramento do peso vivo de suínos por meio da análise de imagens

• Autor(a) principal: GOMES, Nicoly Farias
• Data de Publicação: 2022
• Tipo de documento: Tese
• Idioma: por
• Título da fonte: Biblioteca Digital de Teses e Dissertações da UFRPE
• Texto Completo: http://www.tede2.ufrpe.br:8080/tede2/handle/tede2/9109
• topic: Desempenho animal; Suínos; Análise de imagem; Biometria; Peso; CIENCIAS AGRARIAS::ENGENHARIA AGRICOLA

Determining animal performance requires labor and time to drive the animals to the management center and access the scales for monitoring the animal's live weight, a practice that can result in a stressful step for both the animals and the professionals involved in the process. To overcome the problems inherent to manual weighing, the use of 2D or 3D digital imaging is being employed to estimate or measure the body characteristics of various production animals. In this context, this research was conducted with the objective of developing mathematical models for determining the live weight of pigs from the biometric variables of the animal, programming code and computational procedure to estimate the live weight of pigs in the phases of growth and termination, through the analysis of image obtained by depth camera (3D). The study was conducted at Roçadinho farm, in the Capoeiras district, located in the Agreste region of the state of Pernambuco, where were performed the weighing (kg), manual measurements of five biometric parameters (body length, m; chest circumference, m; (body length, m; chest circumference, m; height at the withers, m; croup height, m; croup width, m) and 3D images by a Kinect® - V2 camera of 20 animals in the growing phase and 24 animals in the termination phase of the crossbreeding between Pietrain and Large White, totaling 44 animals (males and females). The highest determination coefficients (R²) presented by the proposed prediction models for the growing and termination phases were 91.32 and 84.68%, respectively, and the overall model was 98.97%, which denotes the reliability and applicability of the models developed. A Python program was developed to analyze the 3D images. The coefficient of linear determination between estimated weight and volume was 78.5, 74.5 and 97.8% for pigs in the growing, termination and global phases, showing that this relationship is positive and expresses satisfactorily the weight of the animals.