Chen-Yu LiaoJih-Tay HsuTa-Te Lin2024-10-282024-10-282024https://www.scopus.com/record/display.uri?eid=2-s2.0-85206077260&origin=resultslisthttps://scholars.lib.ntu.edu.tw/handle/123456789/722508The integration of a fully automated individual cattle temperature monitoring system is set to bring significant transformations to the operation and management of dairy farming. Early studies indicate that common dairy cow diseases, such as heat stress and milk fever, are preceded by distinct body temperature changes. However, the ineffectiveness of traditional methods makes frequent temperature recording on large commercial farms impractical. Our automated system integrates image processing algorithms, deep learning models, and thermal imaging to continuously record individual dairy cow temperatures, detect abnormalities, and alert farmers for early intervention. We have implemented the YOLOv4 Tiny model for cow face detection for images captured by the thermal imaging camera to facilitate real-time temperature collection, achieving 99% on mAP@0.5 and a processing speed of 25 frames per second (FPS). For dairy cow identity recognition, we utilized a deep learning CNN model with additive angular margin loss (ArcFace) as the loss function. This model, tested on a dataset of 131 cows, achieved a 99% accuracy rate. By integrating cow face recognition and temperature monitoring, we can successfully gather long-term temperature data for each individual cow. For the processing of long-term data, we can compile each cow's average temperature, analyze the relationship between dairy cow body temperature and environmental temperature, and identify any abnormal temperature value. This research demonstrates considerable potential in practical applications of the developed dairy cow temperature monitoring system, highlighting its capability to significantly reduce labor requirements and enhance overall farm management efficiency.falseArcFaceCow Face RecognitionDeep LearningInfrared Temperature MeasurementYOLO[SDGs]SDG2conference paper10.13031/aim.2024003372-s2.0-85206077260