No full-text available
To read the full-text of this research,
you can request a copy directly from the authors.
RF Signal-Based Multipurpose UAV Surveillance System Using Deep Neural Network
... To enhance the detection and classification performance of drone monitoring algorithms, multiple DL models have recently been studied and compared. The predefined networks analysed in [28] reveal that MobileNet V2 is able to outperform SqueezeNet and ResNet in terms of accuracy at the expense of computational complexity. Three custom networks from [29]- [31] are evaluated in [32] on the same dataset against a novel model that is presented and extensively optimised. ...
Unmanned Aerial Vehicles (UAVs), sometimes known as drones, evolved from military to civilian applications, opening up novel perspectives in a variety of everyday services. The rapidly growing consumer interest in amateur drones equipped with high-end cameras compromises the everyday safety and privacy of people. In the literature, a variety of sensing techniques based on different physical phenomena have been proposed for drone detection. Among acoustic, optical, or radar detection systems, passive radiofrequency sensing is the only one that can identify a drone even before it takes off and additionally indicate the operator’s location. A spectrogram-based method is developed and optimised in terms of computing location, resulting in the possibility of sensor grid deployment over a standard Ethernet network. The detection phase involves hardware-accelerated energy sensing to extract the data frames from the background noise. Drone presence is then identified using machine learning based solely on preamble pattern recognition, which reduces the computational effort. The presented procedure is evaluated in an isolated setting employing an open-source dataset and tuned across multiple neural network architectures. Next, the complete sensor processing chain is examined in a real-life scenario. The analytical energy detector stage reaches a margin of roughly -8.7 dB in the signal-to-noise (SNR) ratio. With 1.1 M parameters, the proposed neural network achieves 99.93% simulation accuracy in up to -9.5 dB SNR range. Even after quantization for embedded platform implementation, the device can be used as a stand-alone early intrusion detector or as part of a distributed sensor grid.
Unmanned aerial vehicle (UAV) contributes substan�tial strategic benefits on the Internet-of-military-things (IoMT). However, the untrusted party’s misuse of the UAV may violate the security and even demolish the critical operation in the IoMT system. In addition, data manipulation and falsification using unauthorized access are the significant challenges of the IoMT system. In response to this problem, this study proposes a blockchain integrated convolution neural network (CNN)-based intelligent framework named IoMT-Net for identification and tracking illegal UAV in the IoMT system. Blockchain technology prevents illicit access, data manipulation, and illegal intrusions, as well as stored data on the central control server (CCS). Concurrently, the proposed CNN analyzed the radio frequency
signal sent by the antenna array element to determine the direction of arrival (DoA) for the localization of the illegal UAV. Therefore, a signal model is designed to processes the received signal array through IoMT-Net. Moreover, the proposed CNN model is designed with two different functional modules, such as the resource accuracy trade-off (RAT) module and the unique feature extraction and accuracy boosting (UAB) module, by adopting depth-wise and grouped convolution layers. These sparsely connected convolution layers offer high DoA estimation accuracy while maintaining the network more lightweight. In addition, the skip connection is also leveraged into the sub-units of RAT and UAB modules for sharing features and handling the vanishing gradients problem. Based on the simulation results,
the proposed network achieves superior DoA estimation accuracy
(approximately 97.63% accuracy at 10 dB SNR) and outperforms
other state-of-the-art models.
The surging proliferation in the deployment of unmanned aerial vehicles (UAVs) in various domains has resulted into unsolicited intrusion into private properties and protected areas thereby posing threat to national security. This paper proposed an adaptive scenario-based approach for detecting drone invasion using enhanced YOLOv5 deep learning model to detect different drones and identify attached objects operating under any environment, size, speed, or shape. The dataset consists of 6 drone models and 8 attached weapons manually generated and preprocessed to form samples. In terms of accuracy, sensitivity, and timeliness, the result shows that our model achieved superior detection precision of 100%, sensitivity of 99.9%, F1-score of 87.2% for weapons identification at a shorter time of 0. 021s than other models. The high detection accuracy undoubtedly makes our model well suited for real-time drone monitoring and countering of illegal drones in military offensives with minimal resource usage.
In the past few years, the commercial use of drones has exploded, since they are a safe and cost‐effective solution for many kinds of problems. However, this fact also opens the door for malicious use. This work presents a novel system able to detect and recognise drones from other targets, allowing the police and security agencies to deal with this new aerial thread. The proposed system only uses a persistent range–Doppler radar, avoiding the restrictions of the optical sensors, usually required for the recognition part. The processing is based on constant false alarm rate detection stage, followed by a convolutional neural network that performs the recognition. This network takes as input raw range–Doppler radar data and predicts their class (car, person, or drone). For this purpose, an extensive controlled trial test campaign has been performed, resulting in a novel dataset with more than 17,000 samples of drones, cars, and people, acquired in real outdoor scenarios. As far as authors’ knowledge, this is the first range–Doppler radar database for the recognition of drones and other targets. The high‐accuracy results (99.48%) suggest that this system could be successfully used in security and defence applications to discriminate between drones and other entities.
Recently, Unmanned Aerial Vehicles (UAV)'s, also known as drone, are becoming rapidly popular due to the advancement of their technology and the significant decrease in their cost. Although, commercial drones have proven their effectiveness in many applications such as cinematography, agriculture monitoring, search and rescue, building 3D inspection, and many other day to day applications, they have also been alarmingly used in malicious activities that are targeting to harm individuals and societies which raises great privacy, safety and security concerns. Hence, in this research, we propose a new solution drone detection solution based on the radio frequency emitted during the live communication session between the drone and its controller using the convolutional neural network (CNN). The results of the study have proven the effectiveness of using CNN for drone detection with accuracy and F1 score of over 99.7% and drone identification with accuracy and F1 score of 88.4%. Moreover, the results yielded from this experiment have outperformed those reported in the literature for Radio Frequency (RF) based drone detection using Deep Neural Networks.
This paper investigates the problem of detection and classification of unmanned aerial vehicles (UAVs) in the presence of wireless interference signals using a passive radio frequency (RF) surveillance system. The system uses a multistage detector to distinguish signals transmitted by a UAV controller from the background noise and interference signals. First, RF signals from any source are detected using a Markov models-based naïve Bayes decision mechanism. When the receiver operates at a signal-to-noise ratio (SNR) of 10 dB, and the threshold, which defines the states of the models, is set at a level 3.5 times the standard deviation of the preprocessed noise data, a detection accuracy of 99.8% with a false alarm rate of 2.8% is achieved. Second, signals from Wi-Fi and Bluetooth emitters, if present, are detected based on the bandwidth and modulation features of the detected RF signal. Once the input signal is identified as a UAV controller signal, it is classified using machine learning (ML) techniques. Fifteen statistical features extracted from the energy transients of the UAV controller signals are fed to neighborhood component analysis (NCA), and the three most significant features are selected. The performance of the NCA and five different ML classifiers are studied for 15 different types of UAV controllers. A classification accuracy of 98.13% is achieved by k-nearest neighbor classifier at 25 dB SNR. Classification performance is also investigated at different SNR levels and for a set of 17 UAV controllers which includes two pairs from the same UAV controller models.
Modern technology has pushed us into the information age, making it easier to generate and record vast quantities of new data. Datasets can help in analyzing the situation to give a better understanding, and more importantly, decision making. Consequently, datasets, and uses to which they can be put, have become increasingly valuable commodities. This article describes the DroneRF dataset: a radio frequency (RF) based dataset of drones functioning in different modes, including off, on and connected, hovering, flying, and video recording. The dataset contains recordings of RF activities, composed of 227 recorded segments collected from 3 different drones, as well as recordings of background RF activities with no drones. The data has been collected by RF receivers that intercepts the drone's communications with the flight control module. The receivers are connected to two laptops, via PCIe cables, that runs a program responsible for fetching, processing and storing the sensed RF data in a database. An example of how this dataset can be interpreted and used can be found in the related research article "RF-based drone detection and identification using deep learning approaches: an initiative towards a large open source drone database" (Al-Sa'd et al., 2019).
This paper presents the use of micro-Doppler signatures collected by a multistatic radar to detect and discriminate between micro-drones hovering and flying while carrying different payloads, which may be an indication of unusual or potentially hostile activities. Different features have been extracted and tested, namely features related to the Radar Cross Section of the micro-drones, as well as the Singular Value Decomposition (SVD) and centroid of the micro-Doppler signatures. In particular, the added benefit of using multistatic information in comparison with conventional radar is quantified. Classification performance when identifying the weight of the payload that the drone was carrying while hovering was found to be consistently above 96% using the centroid-based features and multistatic information. For the non-hovering scenarios classification results with accuracy above 95% were also demonstrated in preliminary tests in discriminating between three different payload weights.
In recent years, the availability of commercial unmanned air vehicles (UAVs) has increased enormously because of device miniaturization and low cost. However, the abuse of UAVs needs to be investigated to prevent serious security threats for civilians. Therefore, this paper presents a convolutional neural network-based surveillance system for drone detection and its type identification, namely CNN-SSDI. The network architecture is cleverly designed based on deep convolution layers to successfully learn all intrinsic feature maps of radio-frequency signals that are collected from three different types of drones. Further, a detailed comparative analysis of various kernel impairments of the convolution layer structure was investigated under various performance metrics evaluation and higher accuracy in drone surveillance systems. According to the empirical results, CNN-SSDI can detect a UAV with 99.8% accuracy and recognize drone types with an accuracy of 94.5%, which outperforms other existing drone detection and identification techniques.
This paper presents a convolution neural network (CNN)-based direction of arrival (DOA) estimation method for radio frequency (RF) signals acquired by a nonuniform linear antenna array (NULA) in unmanned aerial vehicle (UAV) localization systems. The proposed deep CNN, namely RFDOA-Net, is designed with three primary processing modules, such as collective feature extraction, multi-scaling feature processing, and complexity-accuracy trade-off, to learn the multi-scale intrinsic characteristics for multi-class angle classification. In several specific modules, the regular convolutional and grouped convolutional layers are leveraged with different filter sizes to enrich diversified features and reduce network complexity besides adopting residual connection to prevent vanishing gradient. For performance evaluation, we generate a synthetic signal dataset for DOA estimation under the multipath propagation channel with the presence of additive noise, propagation attenuation and delay. In simulations, the effectiveness of RFDOA-Net is investigated comprehensively with various processing modules and antenna configurations. Compared with several state-of-the-art deep learning-based models, RFDOA-Net shows the superiority in terms of accuracy with over 94% accuracy at 5 dB signal-to-noise ratio (SNR) with cost-efficiency.
Recent events of drones flying over city centers, official buildings and nuclear installations stressed the growing threat of uncontrolled drone proliferation and the lack of real countermeasure. Indeed, detecting and tracking them can be difficult with traditional techniques. A system to acoustically detect and track small moving objects, such as drones or ground robots, using acoustic cameras is presented. The described sensor, is completely passive, and composed of a 120-element microphone array and a video camera. The acoustic imaging algorithm determines in real-time the sound power level coming from all directions, using the phase of the sound signals. A tracking algorithm is then able to follow the sound sources. Additionally, a beamforming algorithm selectively extracts the sound coming from each tracked sound source. This extracted sound signal can be used to identify sound signatures and determine the type of object. The described techniques can detect and track any object that produces noise (engines, propellers, tires, etc). It is a good complementary approach to more traditional techniques such as (i) optical and infrared cameras, for which the object may only represent few pixels and may be hidden by the blooming of a bright background, and (ii) radar or other echo-localization techniques, suffering from the weakness of the echo signal coming back to the sensor. The distance of detection depends on the type (frequency range) and volume of the noise emitted by the object, and on the background noise of the environment. Detection range and resilience to background noise were tested in both, laboratory environments and outdoor conditions. It was determined that drones can be tracked up to 160 to 250 meters, depending on their type. Speech extraction was also experimentally investigated: the speech signal of a person being 80 to 100 meters away can be captured with acceptable speech intelligibility.
This paper presents a method for optimally combining pixel information from thermal imaging and visible spectrum colour cameras, for tracking an arbitrarily shaped deformable moving target. The tracking algorithm rapidly re-learns its background models for each camera modality from scratch at every frame. This enables, firstly, automatic adjustment of the relative importance of thermal and visible information in decision making, and, secondly, a degree of “camouflage target” tracking by continuously re-weighting the importance of those parts of the target model that are most distinct from the present background at each frame. Furthermore, this very rapid background adaptation ensures robustness to rapid camera motion. The combination of thermal and visible information is applicable to any target, but particularly useful for people tracking. The method is also important in that it can be readily extended for fusion of data from arbitrarily many imaging modalities.