Uso de drones no monitoramento comportamental de bovinos em sistemas intensivos de produção: aplicações, confiabilidade e limitações.

This article analyzes, through qualitative and descriptive-exploratory bibliographic research, the applications, reliability and limitations of drones for the behavioral monitoring of cattle in intensive production systems. Five publications were examined: three field studies in Brazilian feedlots, one dissertation on computer vision applied to drone images, and one pasture study used as a contrast of context. The comparison considered the behaviors assessed, the capture protocol, the validation method and the limitations stated by the authors. In the study with the most direct statistical validation for behavior, agreement between drone images and direct observation was excellent for feeding behavior and good for standing, lying and mounting, whereas drinking, ruminating and walking showed poor agreement. The other works describe qualitative observations, animal counting and automated postural classification, using different metrics that prevent direct numerical comparison. The results indicate that reliability is higher for long-lasting behaviors with a well-defined posture, and lower for dynamic, short-lived behaviors or those subject to occlusion. The available evidence base is small and concentrated in a few sites and research groups, which limits generalization. It is concluded that drones are a complement to conventional observation rather than a substitute, and that standardized protocols are still lacking.