牛羊采食量智能监测技术研究进展OA

Feed intake in individual cattle and sheep can serve as one of the most fundamental physiological indicators to assess animal health status for precision diets and genetic selection.Conventional monitoring is limited to inducing animal stress,such as manual weighing and indicator techniques,including high labor intensity,substantial operational costs,low temporal resolution,and potential disturbance to natural feeding behaviors.Feeding strategies are often required to fully meet the demands of large-scale livestock production,precision management,and animal welfare suitable for controlled environments.Alternatively,intelligent perception has emerged for real-time and continuous monitoring of individual feed intake in cattle and sheep,including animal wearable sensors,machine vision,Internet of Things,edge computing,and deep learning monitoring,particularly for feeding stations,acoustic sensors,accelerometers,pressure detection,and electrophysiological signal monitors.The feeding process can involve a complex sequence of behaviors,including biting,chewing,swallowing,rumination,rumination chewing,and rumination swallowing,all of which show significant correlations with actual intake volume.Feed mass and volume can be measured to generate characteristic acoustic signatures during mastication.Head movement patterns can induce jaw pressure variations and muscle electrophysiological activities,indicating the identifiable visual feeding.Such multi-modal signatures of behavior can be effectively captured for intelligent intake estimation using advanced data fusion and sensing technologies.Here,a systematic review was proposed for the recent advances in intelligent monitoring technologies for feed intake in cattle and sheep,including the working principles,measurement accuracy,optimal application,and limitations of each technology.Feeding intake stations were determined to measure the feed mass before and after consumption.High accuracy and minimal animal interference were offered,requiring no complex algorithmic modeling.Their high equipment costs and maintenance were primarily confined to housing.Acoustic monitoring was used to detect feeding sounds during chewing and swallowing,indicating compact design,easy installation,and minimal animal stress,yet vulnerable to environmental noise interference.Acceleration monitoring was quantified to determine the head movement kinematics for jaw motion acceleration patterns,thus providing cost-effective monitoring solutions with high accuracy.Pressure monitoring was used to identify feeding behaviors after jaw pressure waveform analysis,indicating strong anti-interference with the intake intensity.Electromyographic monitoring was used to predict the intake using electromyographic signals from jaw muscles,with high accuracy and reliable electrode-skin contact during long-term deployment.Vision monitoring was used to non-invasively detect the feed volume or behavior from image data,with high accuracy and substantial infrastructure investment.Several challenges remained for the transition from research prototypes to reliable,widely adopted operation,including excessive energy consumption and limited battery endurance,particularly problematic in extensive grazing;Animal welfare was obtained to minimize the device-induced stress and behavioral disruption;Generalization was limited over diverse animal breeds,production stages,and environmental conditions;Some difficulties were observed to effectively process and extract actionable insights from multi-source big data.Current technological advancements were clarified for the trade-off performance and priorities.Research directions can prioritize the low-power sensing and edge computing for cross-scenario deployment.The integrated data analytics platforms can also be constructed for whole farm optimization.The finding can provide valuable technical references to promote the precision feeding,genetic breeding,and smart farming in the large-scale production of cattle and sheep.