Perpustakaan Digital ITB

The increasing use of drones introduces notable safety and security challenges to both the public and controlled airspace, especially as their compact size and use of non-metallic materials make them difficult to detect with conventional, high-cost radar systems. This research addresses these concerns by developing a modular and high-performance drone detection system based on computer vision, designed for large-scale implementation. The resulting prototype utilizes an NVIDIA Jetson Orin Nano single-board computer in combination with a multi-camera arrangement, SSD MobileNetv2 on TensorRT and a deep learning approach. The system delivers real-time, 360-degree detection, achieving a precision of 0.93, an average confidence score of 0.60, a frame rate of 8.88 frame per second (FPS), and a display delay of just 0.11 seconds, successfully identifying drones with bounding boxes as small as 80 by 60 pixels. These findings establish a strong foundation for development of robust and efficient anti-drone solutions.