Lancaster University researchers have developed an AI-powered image-processing framework to automatically detect and track elephants in drone footage, offering a new tool for wildlife monitoring and conservation. The system uses pioneering computer vision models to analyse video obtained from drones with minimal human input and determine how elephants move around their environment.

Monitoring large mammals across vast landscapes is a major challenge for conservation teams, because traditional methods, such as ground surveys or manual analysis of aerial imagery, are labour-intensive, expensive and slow to deliver data. This limits how often populations can be surveyed and restricts the behavioural insights that scientists can gather.

Post-graduate student in the Lancaster Environment Centre, Chaim Elchik, and colleagues from the University of Liverpool addressed this by combining an object-detection model (YOLOv11) with a tracking system (BoT-SORT), enhanced through a custom re-identification algorithm that maintains accurate tracking of individuals between camera frames. This allows the system to follow elephants across drone footage, even as they move, change direction or briefly pass behind vegetation. The result is a robust, automated system that turns raw video into usable ecological information.

The system can generate key metrics including individual movement speeds, herd movement patterns and clustering behaviour. These indicators are essential for understanding how elephants use their habitat, how groups respond to threats, and how environmental pressures influence behaviour. By automating these analytical tasks, the approach reduces processing time and allows researchers to work through larger volumes of visual data than possible with other systems.

Although tested on drone footage of elephants, the framework has wider potential applications the researchers say. With appropriate training data, the same approach could be adapted to track other endangered species or support broader environmental monitoring. For example, future versions could be used for counting wildlife populations, monitoring migration, studying habitat use, or spotting changes in ecosystem health. As drones become more widely used in conservation, automated analysis could help organisations run more frequent and more detailed surveys without increasing staff burden.

Mr Elchik said of the study: “Our goal was to reduce the workload of ecologists and conservationists by automating the detection and behavioural analysis of elephants. Traditional manual methods are not only time- and labour-intensive but can also be intrusive for the animals. By streamlining these processes, we hope to make wildlife monitoring more efficient, scalable and less disruptive.”

While further work is planned to improve performance in dense habitats and to explore live video applications, the study demonstrates the feasibility and value of integrating modern AI techniques into conservation monitoring. The researchers argue that deep learning will play a growing role in conservation technology by making monitoring more scalable, objective and repeatable. Automation could allow teams to identify issues earlier and develop interventions quickly. The researchers believe that over time, systems like this could form part of real-time monitoring networks, where drones autonomously scan landscapes and alert rangers to unusual behaviour or emerging threats.

The full paper, published in Drone Systems and Applications can be found here