Effective management of individual dairy cow milk yield plays an important role in extending herd longevity and increasing average annual productivity. This paper proposes a smart monitoring and management system designed to track the daily milk yield of individual dairy cows. The proposed system was built based on the Internet of Things (IoT) technology and integrated deep learning models. The proposed system comprises the following main components: an improved RFID tag with energy-harvesting capability integrated with a triaxial accelerometer for behavior monitoring ("Resting," "Grazing," "Moving," and "Milking") and automatic management of cow identification (ID); an electronic weighing unit that can be interfaced with an IoT Node for recording milk yield data; an RFID reader equipped with a 12 dBi antenna for retrieving ID and behavioral data; and a Raspberry-based device embedded with a pre-trained deep learning model that accurately identifies the cow currently positioned at the milking station and integrates this information with milk yield data for transmission to a central server. A Bidirectional Long Short-Term Memory (Bi-LSTM) deep learning model is employed to train and classify dairy cow behaviors during the milking process based on sensor data and the Received Signal Strength Indicator (RSSI), enabling accurate identification of the target cow among neighboring individuals. The system has been deployed and evaluated in a practical environment at the Tan Tai Loc dairy farm (Soc Trang province, Vietnam). Experimental results showed that the trained model can achieve 98% accuracy in behavior recognition and 99.38% accuracy during three months of real-world deployment at the farm. The experimental results demonstrate that the proposed system has potential for practical application, contributing to improved efficiency, enhanced management, and the advancement of smart dairy farm operations in the future.
Tran et al. (Mon,) studied this question.