Integrating Drone-Based Thermal Imaging and AI Tracking to Monitor Population Dynamics of Elusive Aquatic Megafauna

Abstract

The traditional surveys are not effective in monitoring the population dynamics of elusive aquatic megafauna because of the constraint linked to the traditional methods used in the study, which is labor intensive, geographically limited and ineffective with cryptic or nocturnal species. This paper proposes a non-invasive monitoring system, which involves unmanned aerial systems (UAS) with thermal sensors and AI-based detection and tracking pipeline. It employs the optimized architecture that is based on YOLOv11 to enhance the recognition of partially submerged and distant animals. Experiments of thermal data of dolphins in different sea and light environments showed large detection rates and constant person tracking that allowed the census of the population with a small error of 58 percent error versus counts. Operational evaluation also demonstrates that a thermal UAS survey can integrate the same study area in 1.5-3 hours as opposed to 8-12 hours on a boat-based survey, with the least number of disturbances caused by a wildlife. The findings suggest that the thermal-UAS and AI system is an effective, scalable, and light-weight system to produce the timely population forecasts and habitat-use data to assist in conserving aquatic megafauna in the long-run.