AI-powered detection of sea turtle tracks from drone footage for conservation monitoring

• المؤلف الرئيسي: Foo, Yock Haw
• مؤلفون آخرون: Kim Hie Lim
• التنسيق: Final Year Project
• اللغة: English
• منشور في: Nanyang Technological University 2025
• الموضوعات: Computer and Information Science; Earth and Environmental Sciences; Sea turtle conservation; Oriented bounding box model (YOLO); Turtle tracks; Geospatial visualization; Model selection; Drone flight records; Hyperparameter tuning
• الوصول للمادة أونلاين: https://hdl.handle.net/10356/183507

Monitoring sea turtle nesting activity is crucial for marine conservation, as it helps ensure the survival of these endangered species. However, current methods rely heavily on manual identification of turtle tracks, making the process time-consuming and labor-intensive, especially over extensive coastal areas. To address this challenge, the application of object detection models from the You Only Look Once (YOLO) family, with a focus on oriented bounding box (OBB) models, was evaluated to automate and improve turtle track detection. Unlike traditional bounding boxes, which may fragment elongated tracks, OBBs better align with the shape and orientation of turtle tracks, resulting in improved detection accuracy. YOLO-OBB models outperformed standard YOLO models, increasing precision by 8.92%, recall by 4.44%, and mAP@0.50 by 8.67%. Further comparisons among model sizes revealed that the YOLOv8-OBB small (s) model delivered the best results, improving precision by 3.20%, F1-score by 1.48%, and mAP@0.50 by 1.55% over the base nano (n) model. The final optimized model achieved a precision of 0.876, recall of 0.792, and mAP@0.50 of 0.871, and was integrated with drone flight data to autonomously identify and visualize turtle tracks on geospatial maps using QGIS and Python. The resulting workflow facilitates rapid identification of potential nesting sites and enhances large-scale conservation monitoring efforts. By increasing the speed and accuracy of data collection, this approach supports more efficient monitoring of turtle nesting habitats and contributes to the long-term conservation of these endangered species.