Scholarlink

<p>Recent advancements in drone technology have raised significant security concerns, making the classification of drone types for early warning systems increasingly vital. This paper presents an effective solution combining frequency domain signal transformations with deep learning models to improve drone detection accuracy in the presence of noise from various Wi-Fi and Bluetooth signals. Our approach consists of four main stages: i) RF Signal Acquisition; ii) Energy Detection and Wavelet Transformation; iii) Relevant Signal Identification and Noise Elimination; iv) Area Enhancement and Drone Classification. Results indicate that our method demonstrates strong noise tolerance and optimal performance, achieving 99.71% accuracy for drone detection and 98.86% average F1-score across both testing sets. The GPU implementation processes the task in 0.14 seconds, significantly faster than the CPU implementation at 0.35 seconds, making it suitable for real-time applications.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

An Effective RF-Based Solution For Drone Detection And Recognition Amid Noise, Bluetooth, And Wi-Fi Interference

  • 期刊论文

  • 发表日期：

    2025年6月5日

查看全文
  • Trong Thanh Nguyen,
  • Le Cuong Nguyen,
  • Thi-Thanh-Tan Nguyen

摘要

Recent advancements in drone technology have raised significant security concerns, making the classification of drone types for early warning systems increasingly vital. This paper presents an effective solution combining frequency domain signal transformations with deep learning models to improve drone detection accuracy in the presence of noise from various Wi-Fi and Bluetooth signals. Our approach consists of four main stages: i) RF Signal Acquisition; ii) Energy Detection and Wavelet Transformation; iii) Relevant Signal Identification and Noise Elimination; iv) Area Enhancement and Drone Classification. Results indicate that our method demonstrates strong noise tolerance and optimal performance, achieving 99.71% accuracy for drone detection and 98.86% average F1-score across both testing sets. The GPU implementation processes the task in 0.14 seconds, significantly faster than the CPU implementation at 0.35 seconds, making it suitable for real-time applications.