Real-Time and Accurate Drone Detection in a Video with a Static Background

Area of Science:

• Computer Vision
• Artificial Intelligence
• Robotics

Background:

• Increasing prevalence of drones necessitates advanced security systems.
• Illegal drone usage poses significant security risks to guarded areas.
• Existing drone detection methods face challenges with object similarity and high speeds.

Purpose of the Study:

• To develop a real-time, high-accuracy drone detection system using visual information.
• To address the challenge of distinguishing drones from similar objects like birds and airplanes.
• To create a cost-effective drone detection solution.

Main Methods:

• Implemented a two-stage approach: moving object detection and object classification.
• Utilized background subtraction for initial moving object detection.
• Employed a convolutional neural network (CNN) for classifying detected objects (drone, bird, background).

Main Results:

• Achieved high accuracy in drone detection comparable to existing methods.
• Demonstrated high processing speed suitable for real-time applications.
• Validated the effectiveness of the combined background subtraction and CNN approach.

Conclusions:

• The proposed method offers a viable and efficient solution for real-time drone detection.
• The system's performance is robust but can be affected by moving backgrounds.
• Further research could focus on mitigating the impact of dynamic backgrounds on detection accuracy.