Quantifying the movement, behaviour and environmental context of group-living animals using drones and computer...

Area of Science:

• Ecology
• Ethology
• Computer Science

Background:

• Traditional animal behavior data collection methods (direct observation, biologging) have limitations in spatiotemporal resolution and scope.
• Video analysis for animal behavior is challenging due to data volume and georeferencing difficulties.

Purpose of the Study:

• To develop and demonstrate a new system for high-resolution animal behavior analysis in natural environments.
• To integrate drone imagery with computer vision for automated tracking and environmental modeling.

Main Methods:

• Utilized drone-recorded videos and computer vision algorithms to automatically track animal location and body posture.
• Georeferenced animal positions within contemporaneous 3D landscape models.
• Applied the system to gelada monkeys and African ungulates, tracking multiple individuals and classifying them by species and age-sex.

Main Results:

• Successfully tracked multiple animals simultaneously with high spatiotemporal resolution.
• Estimated individual body postures (poses) and extracted environmental features like topography and animal trails.
• Integrated animal movement and posture data with detailed 3D landscape reconstructions.

Conclusions:

• This approach overcomes limitations of traditional methods by providing rich, georeferenced behavioral data.
• Enables in-depth study of animal sensory ecology and decision-making within their physical and social contexts.
• Opens new avenues for wildlife research using integrated drone and AI technologies.