Artificial Intelligence Thermal Imaging for Monitoring Nocturnal Wildlife

Abstract

Introduction: Reliable conservation and management necessitate recurrent data on nocturnal wildlife communities and the environment that affects their activity. Traditional observations are precise yet sporadic and hard to conduct in large or inaccessible areas. Materials and Methods: The paper offers an assessment using artificial intelligence combined with thermal video, air temperature, vegetation density, moonlight, time, and human disturbance observations. The experimental set consisted of 16,800 labeled thermal images collected at 10 observation sites during 60 survey nights. The information was quality controlled, synchronized temporally, and split into training, validation, and testing subsets. Results and Discussion: The developed approach provided 96.5% accuracy, 96.0% precision, 95.6% recall, and 95.8% F1-score. Crucial animal-environment connections involved vegetation density, ambient temperature, moonlight, proximity to the water, and human disturbances. The method enhanced recognition in diverse field conditions and offered interpretable alerts for verification by ecologists. Conclusion: The findings show that artificial intelligence could be used to expand non-invasive observations as long as the outcomes are linked to the habitat specifics and expert observations.