380 Precision Management of Animals: Image Processing and Computer Vision Applications

Abstract

Abstract Growing populations, rising wealth, and urbanization are translating into increased demand for animal products, which causes a need for rapid intensification of production. With that, economic pressures mount on animal producers, who are forced to increase herds’ sizes in order to be commercially feasible, limiting time of interaction with their animals. In contrast, society is demanding closer attention to the needs of individual animals and their well-being, and reduction of the environmental impact of animal production. Paying closer attention to the animal can not only positively impact animal welfare and health, but also increase the capacity of the producer to increase sustainability while still reaching production needs. That’s when Precision Management of Animals (PMA) becomes necessary: The animals become central to the system and, by automatically interpreting their behavior and physical conditions through principles and technologies of process engineering, it is possible to generate data that feeds real-time monitoring and warning systems for producers, so they can take immediate management actions when needed. This leads to better management choices that are not driven only on profits, but, instead, on the needs of the animals and their care, which leads to a more effective use of resources, including antibiotics, grains, and water; improvement of animal welfare; and a data stream that can help guide new facility designs, and genetic evaluation and selection. Image processing and computer vision are examples of technologies that have been used as non-invasive methods of data collection in swine production with multiple purposes, including weight prediction, water usage, aggressive behaviors recognition, detection of lying patterns to evaluate thermal environment, localization of animals, locomotion assessment, behavior classification, gait assessment, animal measurement, and animal counting. When coupled with novel machine learning models, they seem to be a promising approach to solve the existent problems with PMA systems development.