The increasing activity of unmanned aerial vehicles (UAVs) has intensified the demand for reliable and computationally efficient methods for passive radio-frequency (RF) signal detection. In practical RF monitoring scenarios, the environment is often non-stationary and affected by varying noise conditions. Under such circumstances, classical energy-based detectors are sensitive to noise uncertainty, while more robust approaches, such as cyclostationary analysis, require substantially higher computational resources. This work presents a burst-aware cascade method for UAV RF signal presence detection that explicitly addresses this trade-off. The proposed framework combines fast energy-based screening with temporal burst aggregation, applying spectral correlation function (SCF) analysis selectively and only when sustained signal activity is indicated. Detection is performed on fixed-length RF signal chunks, while additional segment-level duration constraints are used to characterize sustained transmissions. The method is evaluated using the publicly available DroneRF dataset and compared against six baseline detectors, including fixed-threshold energy, wavelet-based, blind cyclostationary, two adaptive energy detector variants, and a lightweight convolutional neural network. Experimental results confirm that chunk-level detection remains difficult for all considered methods. Temporal aggregation across longer intervals yields a substantial improvement: the cascade achieves $P_d$ = 1.000 and AUC = 1.000 at the segment level, matching exhaustive cyclostationary detection while reducing per-segment processing time by a factor of 2.46. An additional result is that burst-level concatenation prior to SCF estimation provides implicit coherent integration, preserving $P_d$ = 1.000 at signal amplitude reductions of up to −20 dB where standalone detectors degrade to $P_d$ = 0.995. Overall, burst-aware cascade architectures offer a practical and interpretable approach to RF-based UAV monitoring, providing a well-grounded compromise between detection reliability and computational efficiency under realistic operating conditions.

The growing deployment of unmanned aerial vehicles (UAVs) in both civilian and military domains has created a pressing need for passive, infrastructure-independent identification methods. This paper presents a UAV type identification system based on radio frequency fingerprinting, wherein hardware-induced signal imperfections unique to each transmitter are exploited as discriminative features. Three feature extraction strategies are implemented and compared: power spectral density, short-time Fourier transform spectral statistics, and wavelet packet decomposition. All representations are derived from the publicly available DroneRF dataset and evaluated using a one-dimensional convolutional neural network under stratified five-fold cross-validation with a batch size of 256. Results demonstrate that wavelet packet decomposition, applied at six decomposition levels with a Meyer wavelet basis, yields 91.51 % classification accuracy and a macro F1-score of 0.92 across four target classes, outperforming both spectral alternatives by 36.74 and 4.68 percentage points respectively. Per-class analysis confirms that this advantage includes perfect precision and recall for the minority class. Comparison with a prior result of 97 % obtained under identical architecture but smaller batch size quantifies the large-batch generalization gap for this feature-architecture combination. The paper identifies feature representation as the dominant factor in classification performance within this architecture.

This paper presents the development and evaluation of a lightweight onboard deep learning module for real-time oil spill detection within a distributed UAV-based maritime environmental monitoring system. The study investigates four object detection architectures under controlled training conditions using both baseline and augmented dataset configurations: three convolutional one-stage detectors (YOLOv8s, YOLO11s, YOLO26s) and one transformer-based model (RF-DETR Nano). Performance was assessed using precision, recall, mAP@50, mAP@50–95, and inference latency on an NVIDIA Tesla T4 GPU. Experimental results show that RF-DETR Nano achieves the highest localization accuracy (mAP@50–95 = 0.863) while maintaining real-time throughput (~ 50 FPS), whereas YOLO26s provides the most favourable efficiency–accuracy trade-off among CNN-based models (mAP@50–95 = 0.806 at ~ 67 FPS). The findings demonstrate that accurate and computationally efficient oil spill detection can be performed directly onboard UAV platforms, enabling reliable autonomous maritime monitoring without dependence on continuous ground-based processing.