CUDM: A Combined UAV Detection Model Based on Video Abnormal Behavior.

Area of Science:

• Computer Vision
• Artificial Intelligence
• Surveillance Technology

Background:

• Unmanned Aerial Vehicles (UAVs) offer significant benefits but also pose public safety risks due to malicious use.
• Effective detection of malicious UAVs is crucial for societal security.
• Existing detection methods may rely on expensive equipment like radar.

Purpose of the Study:

• To propose a novel Combined Unmanned Aerial Vehicle Detection Model (CUDM) for detecting malicious UAVs.
• To enhance traditional object detection by incorporating abnormal behavior analysis.
• To develop a cost-effective UAV detection system using standard surveillance cameras.

Main Methods:

• The CUDM model processes video by first detecting abnormal behavior in image frames to filter out irrelevant data.
• Suspicious targets are then identified in the second stage to determine if they are UAVs.
• A custom UAV dataset was created to validate the model's performance.

Main Results:

• The CUDM model achieves accuracy comparable to state-of-the-art object detection methods.
• The proposed model reduces the processing workload by 32%.
• CUDM demonstrates real-time detection capabilities for malicious UAVs.

Conclusions:

• The Combined Unmanned Aerial Vehicle Detection Model (CUDM) offers an efficient and accurate solution for detecting malicious UAVs.
• The model's reliance on ordinary surveillance equipment makes it a practical and accessible solution.
• CUDM effectively balances detection accuracy with reduced workload and real-time performance.