High-quality milk is consistently sought after, yet ensuring the well-being of dairy cows is becoming more difficult. To address this challenge, researchers have introduced a groundbreaking technique that leverages location information through multi-camera systems to monitor individual cows within a barn. This approach facilitates health monitoring, early illness detection, and gestation oversight, making it particularly suitable for large-scale operations focused on maintaining dairy farm health and ensuring a steady supply of premium milk.
As the number of dairy farmers continues to decline each year, the need for high-quality milk remains strong, resulting in an increase in dairy farming activities. While this trend enhances productivity, it complicates the management of individual cow health. Consequently, health management has emerged as a pressing concern in the dairy sector. Timely identification of health issues, prompt diagnosis, disease prevention, and the maintenance of proper breeding cycles are crucial for ensuring reliable and high-quality milk output.
Though there are intrusive methods, such as mechanical devices attached to cows for health tracking, non-invasive techniques are preferred. These approaches are less stressful for the animals, as they do not necessitate physical attachments, making them more suitable for daily farm operations. Techniques include advanced deep learning algorithms that utilize camera-based monitoring and image analysis. The premise is based on the observation that dairy cows frequently display atypical behaviors and movement patterns due to illness, diseases, the estrus cycle, stress, or anxiety. By monitoring individual movements through cameras — including walking patterns, their visits to feeding stations, and water intake frequency — farmers can evaluate cow behavior, which allows for the early prediction of health problems.
A research team from Tokyo University of Science (TUS), Japan, headed by Assistant Professor Yota Yamamoto from the Department of Information and Computer Technology, Faculty of Engineering, along with researchers Mr. Kazuhiro Akizawa, Mr. Shunpei Aou, and Professor Yukinobu Taniguchi, has created a unique location-based method using a multi-camera system to monitor cows throughout an entire barn. Their research was shared online on December 4, 2024, and will be featured in Volume 229 of Computers and Electronics in Agriculture on February 1, 2025.
The method for tracking dairy cows in barns focuses on location data rather than complex image patterns. According to Dr. Yamamoto, “This represents the first initiative to monitor dairy cows across an entire barn using multi-camera setups. Although prior studies utilized multiple cameras to observe various cattle, each camera typically monitored cows individually, often tracking the same cow as different across various cameras. While some techniques allow for continuous monitoring across cameras, they have been restricted to only two or three cameras covering a limited area of the barn.”
The system depends on overlapping camera views to accurately and continuously track dairy cows as they transition from one camera’s view to another, allowing for seamless tracking across multiple cameras. By meticulously managing the number of cameras and their fields of view, the system reduces the potential issues caused by barriers like walls or pillars, which can disrupt camera overlaps in barns with intricate designs. This strategy effectively addresses common obstacles, such as the variability of cows’ fur patterns and distortions from camera lenses, which often hinder conventional tracking approaches.
In testing with video capturing cows moving closely in a barn, this technique achieved about 90% accuracy in monitoring the cows, indicated by Multi-Object Tracking Accuracy, and approximately 80% in Identification F1 score for recognizing each cow individually. This represents a considerable advancement over traditional techniques, which often faced accuracy challenges, particularly in crowded or complicated barn settings. It also performs effectively across various scenarios, whether cows are moving slowly or standing still, and it successfully tackled the issue of cows lying down by adjusting the cow height parameter to 0.9 meters, which is lower than when cows are standing. This modification enhanced tracking accuracy despite changes in posture.
“This technique allows for optimal management and continuous health surveillance of dairy cows, ensuring high-quality milk production at a reasonable cost,” states Dr. Yamamoto. Looking ahead, the team intends to automate the camera setup process to simplify and expedite the installation of the system in different barn environments. They also aspire to improve the system’s capabilities in detecting dairy cows that may exhibit signs of illness or other health concerns, enabling farmers to monitor and manage the health of their herds more effectively.
