With the proliferation of Unmanned Aerial Vehicles (UAVs) to provide diverse critical services, such as surveillance, disaster management, and medicine delivery, the accurate detection of these small devices and the efficient classification of their flight modes are of paramount importance to guarantee their safe operation in our sky. Among the existing approaches, Radio Frequency (RF) based methods are less affected by complex environmental factors. The similarities between UAV RF signals and the diversity of frequency components make accurate detection and classification a particularly difficult task. To bridge this gap, we propose a joint Feature Engineering Generator (FEG) and Multi-Channel Deep Neural Network (MC-DNN) approach. Specifically, in FEG, data truncation and normalization separate different frequency components, the moving average filter reduces the outliers in the RF signal, and the concatenation fully exploits the details of the dataset. In addition, the multi-channel input in MC-DNN separates multiple frequency components and reduces the interference between them. A novel dataset that contains ten categories of RF signals from three types of UAVs is used to verify the effectiveness. Experiments show that the proposed method outperforms the state-of-the-art UAV detection and classification approaches in terms of 98.4% and F1 score of 98.3%.

Unmanned Aerial Vehicles (UAVs, also called drones) have been widely deployed in our living environments for a range of applications such as healthcare, agriculture, and logistics. Despite their unprecedented advantages, the increased number of UAVs and their growing threats demand highperformance management and emergency control strategies. To accurately detect a UAV's working state including hovering and flying, data collection from Radio Frequency (RF) signals is a key step of these strategies and has thus attracted significant research interest. Deep neural networks (DNNs) have been applied for UAV state detection and shown promising potentials. While existing work mostly focuses on improving the DNN structures, we discover that RF signals' pre-processing before sending them to the classification model is as important as improving the DNN structures. Experiments on a dataset show that, after applying proposed pre-processing methods, the 10-time average accuracy is improved from 46.8% to 91.9%, achieving nearly 50% gain comparing with the benchmark work using the same DNN structure. This work also outperforms the state-of-the-art CNN models, confirming the great potentials of data pre-processing for RF-based UAV state detection.

No abstract

No abstract