Tools to identify Canela-Preta chicken (Gallus gallus domesticus) breed by digital images

ABSTRACT This work was justified by the need to develop an accurate and accessible tool to assist in monitoring and controlling poultry flocks. The main objective of this study was to develop an innovative application, specifically designed to identify Canela-Preta chickens using a smartphone as a platform. The research began with the use of 466 images of 152 Canela - Preta chickens. These images were used to create an identification algorithm. The algorithm architecture was chosen based on experiments and performance evaluations. The YOLO v8 framework and the Random Forest algorithm were used to detect objects and locate chickens in the images. To develop the application, an approach was used in which users captured and sent images in real time to a central server and received animal identification results in response. Statistical analyses of the recognition results revealed an average accuracy rate of 99.4% and an average error rate of 0.04%, which highlights the reliability of the approach. The real-time image capture strategy and integration with a central server facilitated the agility of the identification process. The developed application proved to be a promising tool for the accurate and efficient identification of these animals. The findings indicate that the integration of machine learning and image processing technologies can offer innovative solutions to challenges in identifying animals in diverse contexts.