Simulation-based Approach to Classification of Airborne Drones

Lehmann, Lasse; Dall, Jørgen

doi:10.1109/radarconf2043947.2020.9266405

Cited by 12 publications

(5 citation statements)

References 17 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…To verify the proposed method, the monostatic dynamic RCSs of rotating propellers were calculated and compared with the results obtained from a commercial EM tool. First, the radar operating frequency and pulse repetition frequency were set to 9.65 GHz and 20 kHz, respectively, in reference to existing radar specifications [12][13][14]. Next, we choose DJI's Mavic2 [15], which is one of the most popular drones, as the simulation drone model.…”

Section: Resultsmentioning

confidence: 99%

“…Despite the ex-istence of many such studies, the RCS level-based method still has difficulties in distinguishing drones from airborne organisms [2]. To solve this problem, methods based on the signature extracted from the RCS of a target rather than the simple RCS level have been studied [3][4][5][6][7][8][9][10][11][12][13]. Inverse synthetic aperture radar (ISAR) images of small drones were investigated in [4].…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Dynamic RCS Estimation according to Drone Movement Using the MoM and Far-Field Approximation

Lee

Seo

2021

J Electromagn Eng Sci

View full text Add to dashboard Cite

Micro-Doppler signatures from the rotating propellers of a drone can be utilized to distinguish the drone from clutter or airborne organisms with similar radar cross section (RCS) levels, such as birds and bats. To obtain the micro-Doppler signatures of a drone, calculation or measurement of the electric field scattered from the rotating propellers is essential. In this paper, using the relative angle concept and far-field approximation, we propose a way to rapidly estimate the dynamic RCS of a drone with several propellers according to its movement. In addition, based on the fact that the shape of the propeller does not change even if it rotates, we construct an impedance matrix only once and apply the matrix to the method of moments instead of the iterative process of calculating the impedance matrix and inverse matrix for each rotation angle of the propeller. Finally, by using the Fourier transform of the results from the proposed method, the rotation frequencies of the propellers according to the movement of the drone can be obtained.

show abstract

Section: Resultsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Dynamic RCS Estimation according to Drone Movement Using the MoM and Far-Field Approximation

Lee

Seo

2021

J Electromagn Eng Sci

View full text Add to dashboard Cite

show abstract

“…The ML classifier can be explicitly trained to recognise particular drone types or models [14], [18], [33]. This necessitates the availability of datasets per target label, thus higher training data requirements and/or applying a well-defined pipeline for refining a learnt ATC algorithm, for instance with transfer learning [14].…”

Section: A Target Recognition Beyond Drone Versus Non-dronementioning

confidence: 99%

A Review of Automatic Classification of Drones Using Radar: Key Considerations, Performance Evaluation, and Prospects

Ahmad,

Rogers,

Harman

et al. 2024

IEEE Aerosp. Electron. Syst. Mag.

View full text Add to dashboard Cite

Automatic target classification is a critical capability for non-cooperative drone surveillance radars in several defence and civilian applications. It is accordingly a well-established research field and numerous algorithms exist for recognising targets, including miniature unmanned air systems (i.e., small, mini, micro and nano platforms), from their radar signatures. They have notably benefited from advances in machine learning (e.g., deep neural networks) and are increasingly able to achieve remarkably high accuracies. Such classification results are often captured by standard, generic, object recognition metrics and originate from testing on simulated or real radar measurements of drones under high signal to noise ratios. Hence, it is difficult to assess and benchmark the performance of different classifiers under realistic operational conditions. In this paper, we first outline the key challenges and considerations associated with the automatic classification of miniature drones from radar data. We then present a set of important performance metrics, from an end-user perspective. They are relevant to typical drone surveillance system requirements and constraints. Selected examples from real radar observations are shown for illustrations. We also outline here various emerging approaches and future directions that can produce more robust drone classifiers for radar.

show abstract

“…Accuracy analysis Computational time analysis [11] 15, 25 SL 6 [12] 3.25 SL 2 [13] 5.725 SL 1 [14], [15] 8.75 SL 1 [16], [17] 26-40 SL 10 [18] 8-12 SL 2 [19] 9 SL 9 [20] 8-12 SL 1 [21] 8-10 SL 1 [22] 2.5 ML N/A [23] 10 ML 10 [24] 9.35 DL 5 [25] 8.75 DL 1 [26] N/A DL N/A Our work 15,25 SL, ML, DL 6 the signal-to-noise ratio (SNR). Further analysis is provided by means of the Monte Carlo analysis, with emphasis on low SNR conditions.…”

Section: # Of Uavsmentioning

confidence: 99%

Comparative Analysis of Radar Cross Section Based UAV Classification Techniques

Ezuma¹,

Anjinappa²,

Semkin³

et al. 2021

Preprint

View full text Add to dashboard Cite

This work investigates the problem of unmanned aerial vehicles (UAVs) identification using their radar crosssection (RCS) signature. The RCS of six commercial UAVs are measured at 15 GHz and 25 GHz in an anechoic chamber, for both vertical-vertical and horizontal-horizontal polarization. The RCS signatures are used to train 15 different classification algorithms, each belonging to one of three different categories: statistical learning (SL), machine learning (ML), and deep learning (DL). The study shows that while the classification accuracy of all the algorithms increases with the signal-tonoise ratio (SNR), the ML algorithm achieved better accuracy than the SL and DL algorithms. For example, the classification tree ML achieves an accuracy of 98.66% at 3 dB SNR using the 15 GHz VV-polarized RCS test data from the UAVs. We investigate the classification accuracy using Monte Carlo analysis with the aid of boxplots, confusion matrices, and classification plots. On average, the accuracy of the classification tree ML model performed better than the other algorithms, followed by the Peter Swerling statistical models and the discriminant analysis ML model. In general, the classification accuracy of the ML and SL algorithms outperformed the DL algorithms (Squeezenet, Googlenet, Nasnet, and Resnet 101) considered in the study. Furthermore, the computational time of each algorithm is analyzed. The study concludes that while the SL algorithms achieved good classification accuracy, the computational time was relatively long when compared to the ML and DL algorithms. Also, the study shows that the classification tree achieved the fastest average classification time of about 0.46 ms.

show abstract

Simulation-based Approach to Classification of Airborne Drones

Cited by 12 publications

References 17 publications

Dynamic RCS Estimation according to Drone Movement Using the MoM and Far-Field Approximation

Dynamic RCS Estimation according to Drone Movement Using the MoM and Far-Field Approximation

A Review of Automatic Classification of Drones Using Radar: Key Considerations, Performance Evaluation, and Prospects

Comparative Analysis of Radar Cross Section Based UAV Classification Techniques

Contact Info

Product

Resources

About