Drones in Defense: Real-Time Vision-Based Military Target Surveillance and Tracking

In today’s world, with ongoing conflicts and increasing security challenges, the use of drones in defense has become critical for real-time surveillance, precise targeting, and ensuring personnel safety. This study introduces a novel dataset of drone-captured videos tailored for military surveillance, where carefully selected frames were used to optimize training and testing. We evaluated three YOLO deep learning models—YOLOv5, YOLOv8, and YOLOv11—and their nano and small variants to achieve a balance between detection accuracy and computational efficiency. The top-performing model, YOLOv11s, was further integrated with advanced tracking algorithms, including Byte-Track, BoT-SORT, and DeepSORT, to identify the most reliable combination for real-time tracking. Results show that YOLOv11s, paired with ByteTrack, delivers superior accuracy and efficiency in detecting and tracking enemy assets, enabling autonomous monitoring and engagement with minimal human intervention. This work demonstrates significant advancements in autonomous military surveillance.