Assessing the Feasibility of Identifying Unique Features of Wildlife from A Drone's Perspective

Emerging technologies provide novel ways of assessing ecological systems and species. Drones, also known as unmanned aerial vehicles (UAVs), have emerged as a revolutionary tool in conservation, providing new perspectives and the opportunity for new methods of monitoring elusive species. However, the potential of drones, particularly when combined with modern computational tools, remains largely untapped. In this study, we use a combination of drone technology, computer vision and machine learning approaches to recognise and re-identify individual crocodiles from aerial photos. First, the study focused on the construction and validation of a model capable of detecting crocodiles inside photos while also verifying its robustness for varied applications in realworld scenarios. Second, we investigated the potential of re-identifying individual crocodiles based on unique morphological traits, specifically using posture estimation methods and machine learning approaches such as PCA. The conclusion of these processes enabled us to build the framework for future population monitoring and individual re-identification in the wild. Our findings underlined the revolutionary potential of drone applications in animal conservation but also the necessity for multidisciplinary research to address operational and analytical issues, ensuring the effective protection of ecological diversity and the environment.