In the last decade, the rapid development of UAV (Unmanned Aerial Vehicle) enterprises has given rise to a great number of drones, and a part of them illegal flights. Due to their low threshold of entry into the air, it is difficult to monitor low and slow UAVs which threaten the airspace safety. Determining the behaviour of UAVs is a challenging area of airspace regulation. Nowadays, DL (Deep Learning) has become a key research component of radiation signal recognition in many regulatory applications. In order to facilitate the solution to illegal UAV intrusion into airspace without locating the perpetrators, we apply deep learning methods in radio signal recognition based on a real-world signal dataset. Further the model of CNN (Convolutional Neural Network) was pruned to obtain a lightweight network for efficient signal recognition. In this paper, we have verified the first time the effectiveness of DL in recognizing drone raw signals. Then, we also take a closer look at the performance of lightweight models and compare acan achieve 98.9% recognition accuracy, which is at least 1.59% higher than CNN. Finally, we discuss open issues in this field.
Recognition of UAV Image Transmission Signal via Deep Learning Methods
Lect. Notes Electrical Eng.
International Conference on Guidance, Navigation and Control ; 2022 ; Harbin, China August 05, 2022 - August 07, 2022
2023-01-31
11 pages
Aufsatz/Kapitel (Buch)
Elektronische Ressource
Englisch
Traffic Signal Recognition System Using Deep Learning
IEEE | 2023
|Traffic Signal Recognition Using End-to-End Deep Learning
TIBKAT | 2023
|Deep learning methods for automotive radar signal processing
TIBKAT | 2021
|