Chinese researchers have developed a lightweight AI model designed to recognize cow behavior in video footage captured by quadrupedal robots in pastures, improving herd management and feeding efficiency. This was reported by
China DailyPartner of TV BRICS.
The AI model was developed by the Institute of Agricultural Information, Chinese Academy of Agricultural Sciences. The source states that accurately and quickly detecting typical cow behavior is important for disease diagnosis, estrus monitoring, calving prediction, and overall health assessment.
The present invention can identify multiple behavior patterns in complex environments and is suitable for real-time deployment to mobile robots.
By integrating multiscale attention networks, adaptive decomposition and alignment heads, and other advanced techniques, the system addresses challenges such as illumination variations, motion blur, and occlusion of objects in swarms.
The model accurately recognizes six typical cow behaviors, including feeding, resting, walking, and licking, achieving an optimal balance between detection accuracy and computational efficiency. AI models provide essential technical support for the large-scale development of grazing robots.
Photo: Ahabelal/
iStock
