Real-Time Detection of Unauthorized Drones in Restricted Areas

Abstract

This paper presents an intelligent, vision-based drone detection and alert system designed to automatically
identify and track unauthorized drones in restricted airspace. The proposed solution integrates computer vision, deep
learning, and IoT-based communication for real-time monitoring and alert generation. The system employs a
convolutional neural network (CNN) and YOLO-based object detection algorithm to distinguish drones from birds,
planes, and other airborne objects with high accuracy. The model processes live video streams captured from
surveillance cameras and triggers instant alerts via SMS, email, or mobile notifications when a drone is detected.
The backend is implemented using Flask for model inference and REST API communication, while the frontend
dashboard enables real- time visualization of detection events. The system architecture also supports the addition of
multiple cameras and continuous background operation to ensure uninterrupted surveillance. Experimental
evaluations demonstrate high detection precision, fast response time, and robust performance under diverse lighting
and environmental conditions. This solution provides a scalable, low- cost framework for enhancing airspace
security and situational awareness in sensitive environments such as airports, defense zones, and public gatherings.