Method of Binary Detection of Small Unmanned Aerial Vehicles Denys Bakhtiiarov1, 2, Bohdan Chumachenko2, Oleksandr Lavrynenko2, Serhii Chumachenko2, and Vitalii Kurushkin2 1 State Scientific and Research Institute of Cybersecurity Technologies and Information Protection, 3 Maksym Zaliznyak, Kyiv, 03142, Ukraine 2 National Aviation University, 1 Kosmonavta Komarova ave., Kyiv, 03058, Ukraine Abstract This research offers a method for detecting Small Unmanned Aerial Vehicles (sUAVs) in binary using state-of-the-art technology and signal processing techniques. The proposed method combines machine learning and signal analysis techniques to reliably determine the presence of sUAVs in a particular airspace. Pattern recognition, real-time data processing, and spectral analysis are the three primary phases of the approach. Qualitative characteristics of sUAV signals can be identified by spectral analysis. The system can learn and identify these properties and make judgments regarding the presence or absence of sUAVs thanks to the application of machine learning methods. Furthermore, the system’s ability to recognize common patterns of sUAV activity is improved by the integration of pattern recognition. Processing data in real-time guarantees system responsiveness and lowers the number of false signals. The efficiency of the suggested sUAV detection system is strongly demonstrated by experimental results acquired in a variety of environmental circumstances. This highlights how the system can improve airspace monitoring measures’ effectiveness and safety. Keywords 1 Binary detection, signal processing, spectral analysis, detection accuracy, airspace monitoring. 1. Introduction by processing a set of n amplitude beat signals in an FMCW radar, determined for each probing Let’s consider the task of binary detection of FMCW radio signal over a specific observation small-sized targets against the background noise interval t = [0; T]. of the receiving channel of an active radar system The amplitude of the ith beat signal from the perspective of the statistical hypothesis corresponds to the ith probing FMCW radio testing theory in the presence of interference. pulse in the series [5]. Initially, let’s assume that the movement parameters of the UAV are known [1–3], hence 2. Problem Statement the form of the useful signal is known. Similar tasks were addressed by many authors when The formulation of this task is determined by the developing algorithms to detect signals of a specific nature and technical implementation of known form, which comprise a bunch of received receiving channels in typical ground-based short- radio pulses, against the backdrop of additive range active radars with probing FMCW radio Gaussian noise [4]. In this section, we examine signals. In such contemporary digital radars, for the task of detecting a small-sized moving target each probing FMCW radio signal, signal beat CPITS-2024: Cybersecurity Providing in Information and Telecommunication Systems, February 28, 2024, Kyiv, Ukraine EMAIL: bakhtiiaroff@tks.nau.edu.ua (D. Bakhtiiarov); bohdan.chumachenko@npp.nau.edu.ua (B. Chumachenko); oleksandrlavrynenko@tks.nau.edu.ua (O. Lavrynenko); serhii.chumachenko@npp.nau.edu.ua (S. Chumachenko); vitaliy.kurushkin@npp.nau.edu.ua (V. Kurushkin) ORCID: 0000-0003-3298-4641 (D. Bakhtiiarov); 0000-0002-0354-2206 (B. Chumachenko); 0000-0002-3285-7565 (O. Lavrynenko); 0009- 0003-8755-5286 (S. Chumachenko); 0009-0000-4411-0509 (V. Kurushkin) ©️ 2024 Copyright for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0). CEUR Workshop Proceedings (CEUR-WS.org) CEUR ceur-ws.org Workshop ISSN 1613-0073 Proceedings 312 samples from the output of the analog-digital approximation, we’ll describe the correlation converter are processed [6]. This processing function of the Gaussian random process using involves calculating the Fast Fourier Transform, a delta function. As can be inferred from followed by delineating the amplitude spectrum expression (1), random variations in the to determine the distance to relevant objects [7, background’s RCS lead to a multiplicative 8]. Such objects could be a background surface of transformation of the useful signal A (t ) , a particular type. Hence, a distinctive feature of meaning they act as multiplicative interference. the method discussed in this subsection is the The background’s RCS,  f (b 2  b 2 ) , is processing of a set of amplitude samples of signals reflected from background surfaces or nonlinearly incorporated into the amplitude objects included in a burst of radio pulses. expression of the reflected signal, as described First, let’s determine the probability density of by expression (1) (through the square root). instantaneous signal values at the detector’s Let’s determine the density distribution of input both in the presence and absence of UAVs the square root of the RCS and subsequently [9]. the density distribution of the amplitude of the During the observation time t =  0; T  , such reflected signal. It’s well-known that a nonlinear amplitude variation of the radio signal is transformation of a random variable results in represented by the time function A (t ) . This time the alteration of its distribution law in the function A (t ) contains information about the following way [11]: presence of UAVs. Therefore, we’ll consider d ( B) , W ( B) = W [ X = ( B)]  (2) A (t ) it as the useful signal at the detector’s dB input. Let’s first address the task of detecting where W ( X ) is the normal probability density UAVs for a specific set of model parameters, distribution of the random variable X. described by expression (1). W ( B ) is the sought-after probability density Let’s assume that the detector’s input receives distribution of the random variable A. an additive mixture of the useful signal A (t ) and  ( B ) is the function inverse to the function the noise of the receiving path, which we will B = ( X ) . consider as Gaussian and delta-correlated. As In this context, X =  f (b 2  b 2 ) represents a inferred from the materials presented earlier, the power of the useful signal exceeds the power of stationary random process, B =  f (b 2  b 2 ) , the additive noise of the receiving path by 25–30 and the density distribution of instantaneous dB [6]. Under these conditions, the density RCS values at a certain point in time is distribution of the envelope of the observed radio described by a Gaussian law: signal follows a normal law with an average value 2 1   f − m f  equal to the instantaneous value of the useful 1 −  2   f  signal envelope. Let’s represent the observed W ( f ) = e   , (3) 2 f signal Y(t) as the sum of the useful signal A (t ) and the receiving path noise n(t): where m f is the average RCS of the Y (t ) = A (t ) + n(t ) . (1) background ( m =  f 0 ), in square meters; σ f In [10], it is demonstrated that the effective represents the root mean square deviation of scattering surface of the background during the instantaneous RCS values of the background observation of the useful signal acts as a over the observation time of the useful signal, stationary Gaussian random process. The in square meters [12]. temporal correlation interval of this process is In this expression, to simplify the notation, considerably shorter than the duration of the bi-static angle designations b 2 and b 2 have observed useful signal. As a first 313 been omitted. In general, they are functions of  2  1 B − m f  2 −  time during the observation of the useful signal 2 B 2   f  . (7) W ( B) = e   A (t ) . However, in the context discussed, 2  f variations of the mentioned bi-static angles do The useful signal A (t ) is the product of the not result in changes to the background RCS as described by expression (2). deterministic function f (t ) and the square root The random variations in the background of the random variable  f : RCS  f over the observation time of the useful  1 + k A (t ) + 2  k A (t )  cos  (t )   A (t ) =  f  f (t ) =  f  AOTR (8) signal can be represented as the sum of the It’s known that multiplying a stationary average value  f 0 and the fluctuation random process  f by a non-random time component  f : function f (t ) results in a non-stationary  random process with the same distribution law f = f 0 +f . (4) W (  f ) over the observation [14] interval of The multiplicative nature of interference the useful signal. Note that the multiplier AOTR about the useful signal arises from fluctuations also varies over time as the UAV moves due to in the RCS (Radar Cross-Section) of the changes in distances RAB , RBC , RAC , and the background. However, the average value of the antenna gain factor. We consider these changes background’s RCS doesn’t distort the shape of to be insignificant compared to the influence of the useful signal. the oscillating multiplier f (t ) . In this case, the The probability density distribution of the fluctuating component of the background RCS non-stationary process is the result of the can be written as [13]: product of the useful signal A (t ) and the    2 multiplicative interference  f . The non- 1  f  −    2    stationarity of the random process A (t ) is due W ( f ) = 1 e  f    . (5) 2  to the variability of the variance D(  f ) by f f 2 (t ) times, leading to a change in the scale of Expressing the random variable X through B the probability density distribution W (  f ) . and calculating the derivative, we obtain: The change in dispersion over time, d ( B) described by expression (8), according to the X = ( B) = B 2 , = 2 B . (6) dB law of the useful signal can be written as follows: From this, we derive the sought distribution law of the square root from RCS: 2 1   f − m f  −   f 2   f  DA (t ) = AOTR /  1 + k A (t ) + 2  k A (t )  cos  (t )    e   d f (9) − 2    f The probability distribution density of Regrettably, the integrals in formulas (7)– instantaneous values of the useful signal over (9) cannot be expressed in terms of elementary the observation period can be written as functions and can only be determined by follows [15]: numerical integration. 314 2  A 2 − m f  −  /  1 2  A  2  AOTR  1+ k A ( t ) + 2k A ( t )cos    W ( A , t )  e  f  (10) / A OTR  1 + k A (t ) + 2  k A (t )  cos  (t )   2   f 3. Method of Binary Detection of Where σn is the root mean square deviation of the noise measurements in the receiving system. SUAV The random process A (t ) is independent of The useful signal A (t ) , modulated by the receiving system noise. The probability density function of the instantaneous values of multiplicative interference, is observed against the process Y (t ) from formula (1), represented as the backdrop of additive Gaussian noise in the reception path. The noise correlation interval the sum of two independent random processes, is of the reception path does not exceed one determined by the convolution of the probability microsecond. The duration of the useful signal density of the receiving system noise, described is in the order of seconds [16]. Hence, by expression (11), and the probability density of disregarding the correlated noise samples n(t), the useful signal, described by expression (10), as we regard it as white Gaussian noise with a follows:  zero mean value and a probability distribution density of instantaneous values: W (Y , t ) =  WA (Y − n, t ) Wn (n)dn . (12) − 2 1 n  1 −   Or, taking into account (10) and (11), we e  n  . (11) 2  Wn (n) = 2   n obtain:   2 2   2 2 1  (Y − n ) − m f   n   1  (Y − n ) − m f   n   2 2  −  +    −  +   2   f ( t ) f   n   2   f ( t ) f   n   Y  e  dn −  n  e      dn (13) W (Y , t ) = − − f (t )    f  n The density function described by formula signal Y (t ) at the detector’s input characterizes a (13) of the instantaneous values of the observed non-stationary random process with time- signal Y (t ) at the input of the detector varying variance due to the target’s movement characterizes a non-stationary random process relative to the background surface. with time-varying variance due to the motion of It is known that the detection of small targets the target relative to the background surface [16]. is based on the processing of the observed signal. It is known that the detection of small targets In this case, the samples of such a signal are the is based on the processing of the observed signal. amplitudes of the LFMC radio pulse batch In this case, the samples of such a signal are the observed over the time interval t   0, T  . The amplitudes of LFM radio pulses observed over a optimal algorithm for detecting the useful signal time interval t   0, T  . The optimal signal will be sought based on the minimization of the detection algorithm will be sought based on the average risk, the consideration of which leads to minimum average risk criterion, taking into the determination of a specific expression for the account which leads to the determination of a likelihood ratio [17]. specific expression of the likelihood ratio. W (Y / H1 ) . L(Y ) = (14) The density distribution described by formula W (Y / H 0 ) (13) of the instantaneous values of the observed 315 Hypothesis H₁ corresponds to the case of the signal observed at the detector’s input under UAV’s movement in front of the background the assumption of the validity of the hypothesis surface, where the radio signal reflected from H1 is described by the following expression: the background is modulated by the reflected radio signal from the UAV. The model of the Y (t ) = AOTR /   f (b 2  b 2 )  1 + k A (t ) + 2  k A (t )  cos  (t )  = AOTR /   f (b 2  b 2 )  f (t ) (15) The randomness of the process Y (t ) is caused condition of UAV movement. Assuming that the by fluctuations in the Radar Cross-Section (RCS) correlation interval of RCS fluctuations is much  f (b 2  b 2 ) of the background. The modulation smaller than the duration of the useful signal, we can express the density distribution of sample law Y (t ) is determined by the function f(t), which values Y under the presence of a moving UAV as accounts for the non-stationary nature of the follows: sample distribution density Y under the 2 n  Y 2 −m  f   1 i − 2 f i =1  f ( ti )  n 2 Yi 2 AOTR  /2 W (Y / H1 , t ) =  e   (16) / AOTR  2   f i =1 f (ti ) Hypothesis H₀ corresponds to the case of the validity of the hypothesis H1 is described receiving a radio signal reflected from the by the following expression: background. The model of the signal observed at the detector’s input under the assumption of Y (t ) = AOTR /   f (b 2  b 2 ) = AOTR /   f (b 2  b 2 ) (17) The density distribution of the sample Y pattern to expressions (10) and (11), it is under the assumption of hypothesis H₀, in this expressed as follows: case, is stationary, and, following a similar n  (Yi2 −m f )2 1 n − 2   Yi  e 2 AOTR /2 2 f i=1 W (Y / H 0 ) = / (18) AOTR  2  f i =1 Substituting the conditional probability density functions (16) and (17) into (18), we obtain: 2 n n  Y 2 −m  f   n Y   f (ti )  1 i − 2 2 A/2 2  /    i  e OTR  f i=1    AOTR  2    i =1 f (ti ) L(Y ) =   (19) f n n   n  (Yi2 − m f )2 1 − 2  AOTR  2     2 AOTR /2 2 f i=1  /   Yi  e  f  i =1 The obtained expression describes the desired likelihood ratio for the problem of 316 detecting UAVs in the case of multiplicative certain threshold  0 . We simplify the optimal interaction between a known useful signal and detection algorithm described by formula (19) amplitude fluctuations of the background. through logarithmization. The algorithm for detecting UAVs involves comparing the expression (19), L (Y ) , with a 2 n  m   2  Yi 2  n  n  n  1    i =1   +   f  −  (Yi 2 ) 2 + n  m2 f  + 2  m f   Yi 2  1 −  i =1  ti )  (20) f (ti  ) i =1  f (t ) i   f ( n ln  L(Y ) =  ln  f (ti )  − −1 i =1 i =1   2  AOTR /2  2 f Therefore, the optimal algorithm for UAV between a known useful signal and amplitude detection based on the Bayesian criterion, fluctuations of the background, takes the considering the multiplicative interaction following form [17]: n n Yi 4 n  1   Y −  f (t ) − 2  m   Y  1 − f (t )  i 4 2 f i 2   z 1 (ti ) H z= i =1 i =1 i i =1  i  where   z (ti )  H0 2  AOTR /2  2 f A1 − B1 n −  ln f (ti ) + ln  0 —the modified threshold of the detector, −1 z (ti ) = 2  AOTR   f i =1 /2 2 2 n m  A1 =   f  i =1  . (21)  f (ti )  B1 = n  m2 f In the case of discrete sampling of The structure of the optimal detector for observations Y (t ) at the detector’s input, it detecting a moving UAV under the considered follows from expression (21) that to decide the conditions is depicted in Fig. 1. presence of a signal caused by UAV at the detector’s input, a series of operations involving the summation of nonlinearly transformed samples from the observed realization Y (t ) and the multiplication of the square of the realization Y (t ) with a copy of the expected useful signal, followed by summation of the obtained results and comparison with a threshold, should be performed [16]. A distinctive feature of the modified detector threshold z (ti ) is its time Figure 1: Structural diagram of the optimal detector for UAV with known motion parameters dependency proportional to the expected signal due to the non-stationary nature of the The structural diagram of the optimal detector random process. When the threshold level is does not show the synchronization device exceeded, the presence of the moving UAV is responsible for clocking the detector blocks. confirmed; otherwise, a decision is made about Since the additional phase shift during the its absence [17]. reflection of the radio signal by the target and the background surface is random, it is necessary to add a second quadrature channel to the 317 structural diagram presented in Fig. 2.16, where background” line of sight is unknown. In the the function f (t ) is defined with a phase shift of conditions of a priori uncertainty about these π/2 relative to the initial phase of the f (t ) parameters, the application of known approaches to eliminate this uncertainty function. significantly complicates the above algorithm The detection algorithm for UAV (21) is and the structural diagram of the optimal expedient to implement in the digital processing detector [18]. To obtain a practically block of the amplitude signals received in the implementable UAV detection algorithm, we will pulse sequence. However, for the make a series of simplifications relative to the implementation of this algorithm, including observation model Y (t ) . These simplifications setting the threshold, precise knowledge of parameters such as the coordinates (angular will lead to the implementation of a quasi- position) of the phase center (point) of the optimal detection algorithm. background surface reflection, the values of We will consider the model of the observed current bistatic angles, the three-dimensional input signal of the detector on the interval [0, T] shape of the bistatic RCS of the target and as an additive sum of a non-random useful signal background, is required. Furthermore, the start and Gaussian noise limited to  f b in bandwidth: time of the UAV flight relative to the “radar- Y (t ) = AOTR 1 + k A (t ) + 2  k A (t )  cos  (t )  + n(t ) (22) For such an observation model, the detector For the likelihood ratio (22), the probability design has been explored by numerous authors density in the presence of a signal H1 is [19, 20]. Let’s briefly outline the results of expressed as follows: solving this problem. We will assume that the data observation sampling interval is t = 1 . 2  fb n  (Yi2 −m f )2 1 n − 2   Yi  e 2 AOTR /2 2 f i=1 W (Y / H1 ) = (23) / A OTR  2   f i =1 1 n  2 Instead of comparing it to the threshold of −   1 2  the likelihood ratio or function, we can Wn (n) = e  n . (24) compare the logarithms of expressions (25) or 2   n (26). Thus, we obtain the following decision Where σn is the root mean square deviation rule for the considered detection problem [21]: of the noise samples in the receiving path. T  zH1 (t ) Under these conditions, the expression for 2 N 0 0 z= Y (t ) A(t )dt  H0 . (26) the likelihood ratio will be written in a known  z (t ) manner [19]: t n 2 t n The formula z (t ) = ln  0 + E y represents the −  Ai2  Yi  Ai (25) N0 L(Y ) = e e N0 i=1 N0 i=1 . modified threshold. Therefore, the detection The formula can be alternatively expressed device for a moving small-sized target under as a likelihood ratio functional, which these conditions corresponds to the well- N 0 =  2n  f b represents the spectral power known correlation receiver scheme depicted in density of the noise in the receiving system. Fig. 2. 318 The synchronization device ensures coordinated In this case, the reference signal of the operation between the reference generator and the correlation detector should be time-shifted integrator, facilitating the comparison of its output relative to the observed signal, and a procedure signal z (t ) with the threshold. To ensure the for searching and capturing the useful signal functionality of the correlation detector, it is should be performed. To simplify this necessary to multiply the reference and observed procedure, instead of a correlation detector for signals at coincident time points. However, the the useful signal, its version with matched arrival time of the observed signal is unknown. filters should be used. When the useful signal’s time coincides with the impulse response of the matched filter, the value of the correlation integral will match the amplitude of the output signal of the matched filter. The impulse response of the matched filter h() for the useful signal A(t ) is its mirrored copy, shifted in time by t₀. The structural diagram of the UAV Figure 2: Structural diagram of the correlation detector with known parameters of its motion detector for UAV with known parameters of its using matched filters is shown in Fig. 3. motion and initial phase Figure 3: Structural diagram of the UAV detector using a matched filter with known parameters of its motion and initial phase The reference signal A(t ) has a random initial contains two quadrature generators of reference signals, A0 (t ) and A0s (t ) : c phase due to the reflection of radio waves from the target, underlying surfaces, and background. When radio waves are reflected A0 (t ) = ( A0c ) 2 + ( A0s ) 2 = const. (27) from these objects, an additional phase shift Similarly to how the envelope of a harmonic becomes random. signal expressed through quadrature To eliminate the dependence of the components does not depend on time: reference signal on the influence of random phase shifts during the reflection of radio E0 (t ) = E 2 cos() 2 + E 2 sin() 2 = E . (28) waves, we use a structural scheme of a detector In this case, processing the observed signal for a signal of known shape with a random in the small-sized target detection task will initial phase. In this scheme, the detector involve comparing it to a threshold using the following decision statistic: 2 2     z (t ) =   Y (t ) A0c (t − )d  +   Y (t ) A0s (t − )d  , (29)  −   −  where Y (t ) is the observed realization of the A0s (t ) is the sine component of the reference signal at the input of the detector. signal for the matched filter hs (t ) . A0c (t ) is the cosine component of the The structural diagram of the quadrature reference signal for the matched filter hc (t ) . detector for the UAV with known parameters of its motion using matched filters is shown in Fig. 4 [5]. 319 Figure 4: Structural diagram of the quadrature detector for UAV with known parameters of its motion Technology (2020). doi: 4. Conclusions 10.1109/picst51311.2020.9467886. [4] X. Wu, Q. Xue, An Improved CornerNet- Thus, if the motion parameters of the UAV are Lite Method for Pedestrian Detection of known and the background reflection Unmanned Aerial Vehicle Images, China characteristics are sufficiently stable, the Automation Congress (2021) 2322– detector can be represented by a correlation 2327. doi: 10.1109/CAC53003.2021.972 scheme or a scheme with matched filters. 8245. The structural scheme of the SUAV useful [5] O. Veselska et al., A Wavelet-Based signal detection device for the background radar Steganographic Method for Text Hiding based on parallel matched filters is developed. in an Audio Signal, Sensors 22(15) Based on approximations of the widths of (2022) 5832. doi: 0.3390/s22155832. functions describing the change in the amplitude [6] A. Babaryka, et al., Deep Learning of the matched filter response at the mismatch in Methods Application for Object the parameters of the useful signal, the method Detection Tasks Using Unmanned Aerial and algorithm for calculating the number of Vehicles, IEEE 16th International matched filters of the IBPLA detector for the Conference on Advanced Trends in background radar are obtained. Radioelectronics, Telecommunications and Computer Engineering (2022) 808– References 811. doi: 10.1109/TCSET55632.2022. 9766891. [1] S. Liu, et al., Small Target Detection for [7] Z. Hu, et al., Development and Operation Unmanned Aerial Vehicle Images Based Analysis of Spectrum Monitoring on YOLOv5l, 10th International Subsystem 2.4–2.5 GHz Range, Data- Conference on Information Systems and Centric Business and Applications 48 Computing Technology (2022) 210–214. (2020) 675–709. doi: 10.1007/978-3- doi: 10.1109/ISCTech58360.2022.00042. 030-43070-2_29. [2] V. Astapenya, et al., Analysis of Ways and [8] I. Bogachuk, V. Sokolov, V. Buriachok, Methods of Increasing the Availability of Monitoring Subsystem for Wireless Information in Distributed Information Systems based on Miniature Spectrum Systems, in: 2021 IEEE 8 International th Analyzers, in: 5th International Scientific Conference on Problems of and Practical Conference Problems of Infocommunications, Science and Infocommunications. Science and Technology (2021). doi: Technology (2018) 581–585. doi: 10.1109/picst54195.2021.9772161. 10.1109/INFOCOMMST.2018.8632151. [3] V. Astapenya, et al., Last Mile Technique [9] V. Sokolov, P. Skladannyi, A. Platonenko, for Wireless Delivery System using an Video Channel Suppression Method of Accelerating Lens, in: 2020 IEEE Unmanned Aerial Vehicles, in: IEEE 41st International Conference on Problems of International Conference on Electronics Infocommunications. Science and and Nanotechnology (2022) 473–477. 320 doi: 10.1109/ELNANO54667.2022. Techniques Relating to the Malicious Use 9927105. of Small Unmanned Aerial Systems, IEEE [10] D. Pang, et al., A Novel Spatiotemporal Aerospace and Electronic Systems Saliency Method for Low-Altitude Slow Magazine 37(5) (2022) 14–28. doi: Small Infrared Target Detection, IEEE 10.1109/MAES.2022.3151308. Geoscience and Remote Sensing Letters [18] C. Liu, Y. Ming, J. Zhu, Improving the 19 (2022) 1–5. doi: 10.1109/LGRS.2020. Performance of Seabirds Detection 3048199. Combining Multiple Semantic [11] O. Solomentsev, et al., A Procedure for Segmentation Models, IEEE Failures Diagnostics of Aviation Radio International Geoscience and Remote Equipment, International Conference on Sensing Symposium, Waikoloa (2020) Advanced Computer Information 1608–1611. doi: 10.1109/IGARSS39084. Technologies (2023) 100–103. doi: 2020.9324102. 10.1109/ACIT58437.2023.10275337. [19] J. Ni, et al., Ridge Detection and [12] D. Bakhtiiarov, G. Konakhovych, Perceptual Grouping Based Automatic O. Lavrynenko, Protected System of Counting of Rice Seedlings Using UAV Radio Control of Unmanned Aerial Images, 6th International Conference on Vehicle, 4th International Conference on Image, Vision and Computing (2021) 61– Methods and Systems of Navigation and 65. doi: 10.1109/ICIVC52351.2021. Motion Control (2016) 196–199. doi: 9526939. 10.1109/MSNMC.2016.7783141. [20] H. Fang, et al., Infrared Small UAV Target [13] D. Bakhtiiarov, G. Konakhovych, Detection Based on Residual Image O. Lavrynenko, An Approach to Prediction via Global and Local Dilated Modernization of the Hat and COST 231 Residual Networks, IEEE Geoscience and Model for Improvement of Remote Sensing Letters 19 (2022) 1–5. Electromagnetic Compatibility in doi: 10.1109/LGRS.2021.3085495. Premises for Navigation and Motion [21] Z. Hong, et al., Highway Crack Control Equipment, IEEE 5th Segmentation from Unmanned Aerial International Conference on Methods Vehicle Images Using Deep Learning, and Systems of Navigation and Motion IEEE Geoscience and Remote Sensing Control (2018) 271–274. doi: Letters 19 (2022) 1–5. doi: 10.1109/MSNMC.2018.8576260. 10.1109/LGRS.2021.3129607. [14] J. Yu, et al., Spatial Cognition-Driven Deep Learning for Car Detection in Unmanned Aerial Vehicle Imagery, IEEE Transactions on Cognitive and Developmental Systems 14(4) (2022) 1574–1583. doi: 10.1109/TCDS.2021. 3124764. [15] Y. Zhang, et al., Self-Attention Guidance and Multiscale Feature Fusion-Based UAV Image Object Detection, IEEE Geoscience and Remote Sensing Letters 20 (2023) 1–5. doi: 10.1109/LGRS.2023. 3265995. [16] V. Larin, et al., Prediction of the Final Discharge of the UAV Battery Based on Fuzzy Logic Estimation of Information and Influencing Parameters, IEEE 3rd KhPI Week on Advanced Technology (2022) 1–6. doi: 10.1109/KhPIWeek57 572.2022.9916490. [17] C. Swinney, J. Woods, A Review of Security Incidents and Defence 321