Intelligent Identification and Classification of Small UAV Remote Control Signals Based on Improved Yolov5-7.0
Minjing Li,Donglai Hao,Jiaming Wang,Shuozhe Wang,Zijian Zhong,Zhiwen Zhao
DOI: https://doi.org/10.1109/access.2024.3376738
IF: 3.9
2024-03-27
IEEE Access
Abstract:At present, an increasing number of small UAVs(Unmanned Aerial Vehicles) are commercialized and common, and the application of small UAVs has very good development prospects, such as UAV distribution services, UAV aerial photography services, and UAV formation performance. However, the misuse of small drones poses a significant threat. Lawbreakers use small drones equipped with various sensors to spy on personal privacy, steal corporate secrets, and threaten national conferences, which have had many adverse effects on society. In future wars, these small drones will perhaps be used on the battlefield along with high-end weapons. Therefore, it is necessary to find a solution for effectively identifying the basic information of UAV. For the existence of UAV and various small UAV types, this paper proposes a combination of RF sensing and target detection techniques with target detection algorithms to learn RF signal frequency frequency hopping features to detect UAV presence and identify the detected UAV. First, the RF signal of the UAV was obtained in real time by a software radio, and the time-frequency analysis of the short-time Fourier transform and wavelet transform is performed to generate frequency domain images with retained frequency hopping features. Then, the improved Yolov5-7.0 target detection model was employed for training, and finally, the trained model was used for identification and classification. The results showed that the method can effectively assist the detection and classification of UAVs that were obtained by identifying and classifying 5400 unlabeled images. The F1 score was 0.93, and the three assessment measures of P(Precision), R(Recall), and mAP(mean Average Precision) were 1.00, 1.00, and 0.967, respectively.
computer science, information systems,telecommunications,engineering, electrical & electronic