Pawel Biernacki
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Pawel Biernacki has a total of 9 co-authors. Here are the top co-authors by number of shared publications:
- Urszula Libal
Wroclaw University of Science and Technology, Faculty of Electronics, Photonics and Microsystems, Poland
15 shared publications - Agnieszka Wielgus
Wroclaw University of Science and Technology,Faculty of Electronics, Photonics and Microsystem,Wroclaw,Poland
3 shared publications - Szymon Sołtysiak
Kraina AI, Department of Artificial Intelligence, Faculty of Information and Communication Technology, Wrocław University of Science and Technology, Wrocław, Poland
3 shared publications - Stanislaw Gmyrek
Department of Acoustics, Multimedia and Signal Processing, Faculty of Electronics, Photonics and Microsystems Wroclaw University of Science and Technology Wroclaw Poland
2 shared publications - Adam Stefaniak
Wroclaw University of Science and Technology,Faculty of Electronics, Photonics and Microsystems,Wroclaw,Poland
1 shared publication Dawid Sysak
Department of Acoustics, Multimedia and Signal Processing Wroclaw University of Science and Technology Faculty of Electronics, Photonics and Microsystems Wroclaw Poland
1 shared publication
Latest Publications
- Open Access
- 29 October 2026
- Published by Polish Academy of Sciences Chancellery in International Journal of Electronics and Telecommunications
- Vol. 72 (4) , 1-7
Enhancing the Resilience of CNN-KAN Hybrid Architectures to Environmental Distortions in Traffic Sign Classification via Focused Data Augmentation
This paper investigates the efficacy of integrating Kolmogorov-Arnold Networks (KANs) into deep learning architectures to enhance traffic sign recognition robustness against environmental distortions. We evaluate three architectures—a baseline CNN, a hybrid CNN-KAN, and a fully integrated CKAN—using a stochastic augmentation pipeline comprising rotation, simulated rain, blur, and brightness shifts. Experimental results on the GTSRB dataset demonstrate that the hybrid CNN-KAN achieves superior accuracy on clean dataset. KAN-based models excel at handling geometric transformations when trained with augmentation but exhibit higher sensitivity to signal degradation (heavy blur, low light) compared to standard convolutions. We conclude that while KANs offer sharper decision boundaries, their learnable activation functions require extensive augmentation to match the inherent stability of traditional CNNs in low-signal-to-noise environments.
- 29 October 2026
- Published by Polish Academy of Sciences Chancellery in International Journal of Electronics and Telecommunications
- Vol. 72 (4) , 1-7
Open AccessEnhancing the Resilience of CNN-KAN Hybrid Architectures to Environmental Distortions in Traffic Sign Classification via Focused Data Augmentation
This paper investigates the efficacy of integrating Kolmogorov-Arnold Networks (KANs) into deep learning architectures to enhance traffic sign recognition robustness against environmental distortions. We evaluate three architectures—a baseline CNN, a hybrid CNN-KAN, and a fully integrated CKAN—using a stochastic augmentation pipeline comprising rotation, simulated rain, blur, and brightness shifts. Experimental results on the GTSRB dataset demonstrate that the hybrid CNN-KAN achieves superior accuracy on clean dataset. KAN-based models excel at handling geometric transformations when trained with augmentation but exhibit higher sensitivity to signal degradation (heavy blur, low light) compared to standard convolutions. We conclude that while KANs offer sharper decision boundaries, their learnable activation functions require extensive augmentation to match the inherent stability of traditional CNNs in low-signal-to-noise environments.
- Open Access
- 2 June 2026
- Published by Polish Academy of Sciences Chancellery in International Journal of Electronics and Telecommunications
- Vol. 72 (2) , 1-8
Robustness Analysis of CNN and Convolutional KAN Architectures against Weather and Geometric Distortions in Traffic Sign Recognition
This paper investigates the robustness of traffic signclassification models against real-world visual disturbances. Weconduct a comparative evaluation of three distinct architectures:a standard CNN, a hybrid CNN enhanced with Kolmogorov-Arnold dense layers (CNN-KAN), and a fully convolutionalKolmogorov-Arnold Network (CKAN). Unlike traditional CNNs,the KA-based models utilize learnable activation functions, potentiallyoffering improved resilience. The experiments were conductedusing the German Traffic Sign Recognition Benchmark(GTSRB) dataset, containing 43 classes of traffic signs. Modelswere trained and tested on both original images and versionsdegraded by controlled disturbances, including rotation, blur,brightness variation, and simulated rain. The results demonstratethat the proposed CNN-KAN model provides consistently superiorperformance under small-to-moderate rotations (up to 20degrees) and moderate brightness increases, achieving the highestaccuracy in all rain-mask scenarios. It remains competitiveunder blur, where it ranks second only to the standard CNN.Performance decreases were observed only at extreme brightnesslevels, where both the standard CNN and CKAN maintainedhigher stability. Overall, the findings highlight the potential ofKolmogorov-Arnold-based architectures for improving robustnessin traffic sign recognition systems operating under realisticand dynamically changing environmental conditions.
- 2 June 2026
- Published by Polish Academy of Sciences Chancellery in International Journal of Electronics and Telecommunications
- Vol. 72 (2) , 1-8
Open AccessRobustness Analysis of CNN and Convolutional KAN Architectures against Weather and Geometric Distortions in Traffic Sign Recognition
This paper investigates the robustness of traffic signclassification models against real-world visual disturbances. Weconduct a comparative evaluation of three distinct architectures:a standard CNN, a hybrid CNN enhanced with Kolmogorov-Arnold dense layers (CNN-KAN), and a fully convolutionalKolmogorov-Arnold Network (CKAN). Unlike traditional CNNs,the KA-based models utilize learnable activation functions, potentiallyoffering improved resilience. The experiments were conductedusing the German Traffic Sign Recognition Benchmark(GTSRB) dataset, containing 43 classes of traffic signs. Modelswere trained and tested on both original images and versionsdegraded by controlled disturbances, including rotation, blur,brightness variation, and simulated rain. The results demonstratethat the proposed CNN-KAN model provides consistently superiorperformance under small-to-moderate rotations (up to 20degrees) and moderate brightness increases, achieving the highestaccuracy in all rain-mask scenarios. It remains competitiveunder blur, where it ranks second only to the standard CNN.Performance decreases were observed only at extreme brightnesslevels, where both the standard CNN and CKAN maintainedhigher stability. Overall, the findings highlight the potential ofKolmogorov-Arnold-based architectures for improving robustnessin traffic sign recognition systems operating under realisticand dynamically changing environmental conditions.
- 18 May 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 267-272
Sub-Pixel Localization of Radar Targets in Spatially Correlated Clutter
Precise localization of radar targets is a fundamental requirement in many applications, including geodetic calibration, infrastructure monitoring and military target detection. This paper presents a comprehensive hierarchical framework for the automatic detection and sub-pixel localization of diverse target geometries within spatially correlated clutter. We propose a two-stage methodology: first, a Constant False Alarm Rate (CFAR) adaptive thresholding algorithm identifies candidate Regions of Interest (RoI). Second, we evaluate two localization cases: 1) the high-precision positioning of isolated corner reflectors, and 2) the center-of-gravity estimation for larger extended targets such as ships or buildings. The sub-pixel target centers are refined using competing estimators, including Center of Gravity, Parabolic Interpolation, and Non-Linear Least Squares fitting to Gaussian Model. Numerical experiments demonstrate that while algebraic estimators suffer from significant bias in correlated clutter, the Gaussian Model Fitting method achieves precision approaching the Cramer-Rao Lower Bound. Furthermore, we show that for the cases of overlapping responses for extended structures, SuperGaussian modeling are required to prevent localization collapse. The proposed workflow provides a robust, unified solution for automated target analysis in complex radiometric environments.
- 18 May 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 267-272
Sub-Pixel Localization of Radar Targets in Spatially Correlated Clutter
Precise localization of radar targets is a fundamental requirement in many applications, including geodetic calibration, infrastructure monitoring and military target detection. This paper presents a comprehensive hierarchical framework for the automatic detection and sub-pixel localization of diverse target geometries within spatially correlated clutter. We propose a two-stage methodology: first, a Constant False Alarm Rate (CFAR) adaptive thresholding algorithm identifies candidate Regions of Interest (RoI). Second, we evaluate two localization cases: 1) the high-precision positioning of isolated corner reflectors, and 2) the center-of-gravity estimation for larger extended targets such as ships or buildings. The sub-pixel target centers are refined using competing estimators, including Center of Gravity, Parabolic Interpolation, and Non-Linear Least Squares fitting to Gaussian Model. Numerical experiments demonstrate that while algebraic estimators suffer from significant bias in correlated clutter, the Gaussian Model Fitting method achieves precision approaching the Cramer-Rao Lower Bound. Furthermore, we show that for the cases of overlapping responses for extended structures, SuperGaussian modeling are required to prevent localization collapse. The proposed workflow provides a robust, unified solution for automated target analysis in complex radiometric environments.
- 18 May 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-6
Comparison of Radar-Camera Fusion Strategies for Road Traffic Users Detection
This paper presents an approach to environmental perception for Advanced Driver Assistance Systems (ADAS) and autonomous vehicles by classifying road traffic participants using a combination of radar micro-Doppler signatures and RGB camera imagery. The primary contribution involves the development and validation of a deep learning-based classification framework integrated with a specialized digital signal processing (DSP) pipeline. The methodology evaluates deep convolutional neural network (CNN) architectures, including MobileNetV2 and EfficientNetB0, within a unimodal and multimodal context. Two distinct fusion strategies are investigated: Early Fusion, which concatenates RGB data with radar spectrograms into a four-channel input tensor, and Late Fusion, which leverages independent deep network backbones in a parallel dual-stream architecture where deep features are extracted independently from each modality before integration. Furthermore, the study explores autoencoder-based feature extraction, demonstrating that a Supervised Encoder Classifier (AEC) significantly outperforms unsupervised approaches. Experimental results indicate that for lightweight architectures, the proposed AEC-based models achieve superior classification accuracy compared to baseline CNNs. These findings highlight the efficacy of the proposed methodology for real-time deployment in resource-constrained automotive edge-computing environments.
- 18 May 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-6
Comparison of Radar-Camera Fusion Strategies for Road Traffic Users Detection
This paper presents an approach to environmental perception for Advanced Driver Assistance Systems (ADAS) and autonomous vehicles by classifying road traffic participants using a combination of radar micro-Doppler signatures and RGB camera imagery. The primary contribution involves the development and validation of a deep learning-based classification framework integrated with a specialized digital signal processing (DSP) pipeline. The methodology evaluates deep convolutional neural network (CNN) architectures, including MobileNetV2 and EfficientNetB0, within a unimodal and multimodal context. Two distinct fusion strategies are investigated: Early Fusion, which concatenates RGB data with radar spectrograms into a four-channel input tensor, and Late Fusion, which leverages independent deep network backbones in a parallel dual-stream architecture where deep features are extracted independently from each modality before integration. Furthermore, the study explores autoencoder-based feature extraction, demonstrating that a Supervised Encoder Classifier (AEC) significantly outperforms unsupervised approaches. Experimental results indicate that for lightweight architectures, the proposed AEC-based models achieve superior classification accuracy compared to baseline CNNs. These findings highlight the efficacy of the proposed methodology for real-time deployment in resource-constrained automotive edge-computing environments.
- 8 July 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 153-158
Parametric and Geometric Active Contour Models for Image Segmentation
Image segmentation, a fundamental task in image processing, aims to partition an image into semantically meaningful regions corresponding to distinct objects or areas. Active contour models have gained significant attention due to their ability to accurately delineate object boundaries. This paper presents an implementation and evaluation of selected active contour algorithms for image segmentation. Specifically, we focus on two prominent approaches: the parametric Snake model with a contour explicitly described by a curve, and the geometric Chan-Vese model based on level set approach. Our objective is to assess the performance of these algorithms on images with varying levels of complexity and compare our implementations. Both synthetic images with well-defined boundaries and realworld images with complex structures and noisy backgrounds were considered. To evaluate segmentation quality, we employed objective metrics such as Intersection over Union (IoU) and F1-score, for detailed pixel-level analysis. Our experimental results highlight the strengths and limitations of each approach, providing insights into their suitability for different image segmentation tasks. The findings can serve as a foundation for further optimization and adaptation of these algorithms to more complex image segmentation challenges.
- 8 July 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 153-158
Parametric and Geometric Active Contour Models for Image Segmentation
Image segmentation, a fundamental task in image processing, aims to partition an image into semantically meaningful regions corresponding to distinct objects or areas. Active contour models have gained significant attention due to their ability to accurately delineate object boundaries. This paper presents an implementation and evaluation of selected active contour algorithms for image segmentation. Specifically, we focus on two prominent approaches: the parametric Snake model with a contour explicitly described by a curve, and the geometric Chan-Vese model based on level set approach. Our objective is to assess the performance of these algorithms on images with varying levels of complexity and compare our implementations. Both synthetic images with well-defined boundaries and realworld images with complex structures and noisy backgrounds were considered. To evaluate segmentation quality, we employed objective metrics such as Intersection over Union (IoU) and F1-score, for detailed pixel-level analysis. Our experimental results highlight the strengths and limitations of each approach, providing insights into their suitability for different image segmentation tasks. The findings can serve as a foundation for further optimization and adaptation of these algorithms to more complex image segmentation challenges.
- 8 July 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 147-152
Traffic Sign Recognition with Convolutional Kolmogorov-Arnold Networks
This paper investigates traffic sign classification, a well-established problem within the field of image recognition. Traffic sign classification plays a crucial role in advanced driverassistance systems, traffic monitoring systems, and autonomous vehicles, which are rapidly gaining prominence in modern transportation. However, accurately classifying traffic signs presents significant challenges due to the inherent variability in sign designs, varying lighting conditions, and the presence of adverse weather conditions. This study explores the application of convolutional neural networks (CNNs) for traffic sign recognition, comparing three distinct architectures: 1) a standard CNN, 2) a CNN with Kolmogorov-Arnold dense layers (CNNKAN), and 3) a convolutional network incorporating Kolmogorov-Arnold layers within the convolutional stage (CKAN). CNNs excel in image analysis due to their ability to effectively extract relevant features. Unlike CNNs where activation functions (like ReLU or sigmoid) are fixed, KA networks offer the advantage of learnable activation functions on the connections between neurons. The experiments were conducted on the German Traffic Sign Recognition Benchmark (GTSRB) dataset, comprising 43 classes of German traffic signs with $32 \times 32$ pixel images. To evaluate the robustness of each architecture, we employed various image distortion techniques, including rotations, blurring, brightness adjustments, and simulated weather effects such as rain. These interferences were introduced to assess the performance of each model under challenging real-world conditions.- 8 July 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 147-152
Traffic Sign Recognition with Convolutional Kolmogorov-Arnold Networks
This paper investigates traffic sign classification, a well-established problem within the field of image recognition. Traffic sign classification plays a crucial role in advanced driverassistance systems, traffic monitoring systems, and autonomous vehicles, which are rapidly gaining prominence in modern transportation. However, accurately classifying traffic signs presents significant challenges due to the inherent variability in sign designs, varying lighting conditions, and the presence of adverse weather conditions. This study explores the application of convolutional neural networks (CNNs) for traffic sign recognition, comparing three distinct architectures: 1) a standard CNN, 2) a CNN with Kolmogorov-Arnold dense layers (CNNKAN), and 3) a convolutional network incorporating Kolmogorov-Arnold layers within the convolutional stage (CKAN). CNNs excel in image analysis due to their ability to effectively extract relevant features. Unlike CNNs where activation functions (like ReLU or sigmoid) are fixed, KA networks offer the advantage of learnable activation functions on the connections between neurons. The experiments were conducted on the German Traffic Sign Recognition Benchmark (GTSRB) dataset, comprising 43 classes of German traffic signs with $32 \times 32$ pixel images. To evaluate the robustness of each architecture, we employed various image distortion techniques, including rotations, blurring, brightness adjustments, and simulated weather effects such as rain. These interferences were introduced to assess the performance of each model under challenging real-world conditions.- 7 April 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Signal Processing Letters
- Vol. 32 (10709908) , 1665-1669
Nonlinear Schur-Type Audio Signal Parameterization for Convolutional Networks
This article introduces a novel signal parameterization approach, termed nonlinear Schur-type signal parameterization, designed to enhance machine learning tasks such as signal classification and recognition. Traditional linear parameterization methods often struggle with the complex, nonlinear nature of real-world data. The mathematical foundation of the proposed parameterization method is extraction of Schur coefficients. The presented method is scalable and can be adjusted to the signal nature. The nonlinear Schur parameterization produces a matrix of Schur coefficients in time, dedicated to be a 2D input of convolutional neural networks (CNN). The performed experiments for the audio signals from open access datasets show that the signal representation in the form of the Schur coefficients is very efficient for recognition performance. The results obtained by CNN show an improvement in the classification accuracy in comparison with solutions based on preprocessing in frequency domain as FFT or MFCC.
- 7 April 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Signal Processing Letters
- Vol. 32 (10709908) , 1665-1669
Nonlinear Schur-Type Audio Signal Parameterization for Convolutional Networks
This article introduces a novel signal parameterization approach, termed nonlinear Schur-type signal parameterization, designed to enhance machine learning tasks such as signal classification and recognition. Traditional linear parameterization methods often struggle with the complex, nonlinear nature of real-world data. The mathematical foundation of the proposed parameterization method is extraction of Schur coefficients. The presented method is scalable and can be adjusted to the signal nature. The nonlinear Schur parameterization produces a matrix of Schur coefficients in time, dedicated to be a 2D input of convolutional neural networks (CNN). The performed experiments for the audio signals from open access datasets show that the signal representation in the form of the Schur coefficients is very efficient for recognition performance. The results obtained by CNN show an improvement in the classification accuracy in comparison with solutions based on preprocessing in frequency domain as FFT or MFCC.
- Open Access
- 7 January 2025
- Published by Polish Academy of Sciences Chancellery in International Journal of Electronics and Telecommunications
- p. 413-418
Fuel Level Estimation in Tank of Truck in Motion
The paper presents the results of a case study on estimating the fuel level in the tank of a motor vehicle. A method based on the concept of particle filtering of noisy measurement data is proposed. The algorithm designed using the Sequential Monte Carlo method with Sequential Importance Sampling is combined with classical digital filters used for signal filtering. In the simulations, real data obtained by measuring fuel levels in the tanks of TIR heavy trucks from one of the Polish trucking companies are used. The performance of the applied method was considered in various measurement situations, such as refueling, driving on an uneven road surface, driving on steep roads, and fading of the measurement signals.
- 7 January 2025
- Published by Polish Academy of Sciences Chancellery in International Journal of Electronics and Telecommunications
- p. 413-418
Open AccessFuel Level Estimation in Tank of Truck in Motion
The paper presents the results of a case study on estimating the fuel level in the tank of a motor vehicle. A method based on the concept of particle filtering of noisy measurement data is proposed. The algorithm designed using the Sequential Monte Carlo method with Sequential Importance Sampling is combined with classical digital filters used for signal filtering. In the simulations, real data obtained by measuring fuel levels in the tanks of TIR heavy trucks from one of the Polish trucking companies are used. The performance of the applied method was considered in various measurement situations, such as refueling, driving on an uneven road surface, driving on steep roads, and fading of the measurement signals.
- Open Access
- 1 January 2025
- Published by Institution of Engineering and Technology (IET) in IET Radar, Sonar & Navigation
- Vol. 19 (1)
Monte Carlo Modelling of Echoes Reflected by High‐Rise Architectural Landmarks in UAV Anticollision Radar
This paper presents a novel approach to synthesising radar echoes for unmanned aerial vehicle (UAV) anticollision systems, specifically focusing on the challenges posed by high‐rise architectural landmarks in urban environments. We employ a Monte Carlo method to generate synthetic radar data that accurately reflects the statistical properties of real‐world radar echoes, derived from data collected using a custom‐designed X‐band radar. Our methodology involves the probabilistic modelling of radar echoes for three distinct classes: large‐scale arena building, sky‐scraping slender spire and background noise, using kernel density estimation (KDE). This approach allows for the creation of a large database of synthetic radar signatures essential for training and validating machine learning algorithms intended for use in UAV collision avoidance systems. The results demonstrate the efficacy of our method in preserving the statistical characteristics of real radar data, enabling the generation of high‐fidelity synthetic echoes that can significantly enhance the development and testing of UAV navigation and obstacle avoidance systems.
- 1 January 2025
- Published by Institution of Engineering and Technology (IET) in IET Radar, Sonar & Navigation
- Vol. 19 (1)
Open AccessMonte Carlo Modelling of Echoes Reflected by High‐Rise Architectural Landmarks in UAV Anticollision Radar
This paper presents a novel approach to synthesising radar echoes for unmanned aerial vehicle (UAV) anticollision systems, specifically focusing on the challenges posed by high‐rise architectural landmarks in urban environments. We employ a Monte Carlo method to generate synthetic radar data that accurately reflects the statistical properties of real‐world radar echoes, derived from data collected using a custom‐designed X‐band radar. Our methodology involves the probabilistic modelling of radar echoes for three distinct classes: large‐scale arena building, sky‐scraping slender spire and background noise, using kernel density estimation (KDE). This approach allows for the creation of a large database of synthetic radar signatures essential for training and validating machine learning algorithms intended for use in UAV collision avoidance systems. The results demonstrate the efficacy of our method in preserving the statistical characteristics of real radar data, enabling the generation of high‐fidelity synthetic echoes that can significantly enhance the development and testing of UAV navigation and obstacle avoidance systems.
- Open Access
- 1 January 2025
- Published by Institution of Engineering and Technology (IET) in IET Radar, Sonar & Navigation
- Vol. 19 (1)
High‐Rise Architectural Landmarks Detection and Identification by Spatio‐Probabilistic Models for UAV Anti‐Collision Radar—A Real Data Case
Unmanned aerial vehicles (UAVs) heavily rely on GPS, a system vulnerable to signal interference in complex urban environments. Although radar systems offer a robust alternative due to their ability to effectively penetrate adverse weather and operate in darkness, a key challenge remains: reliably identifying static architectural landmarks from sparse and noisy radar echoes. This paper proposes a novel method for creating spatio‐probabilistic models (SPMs) of radar echoes from high‐rise urban landmarks, enabling independent, radar‐based UAV localisation. We employ kernel density estimation on real radar data, acquired with a custom‐designed X‐band ENAVI radar, focusing on large arena buildings and slender spires. These SPMs are then used to detect and identify landmarks by calculating the divergence between the probability distributions of the real‐time received echoes and the preestimated reference models. Our evaluation, using probabilistic divergence metrics on Wrocław's Centennial Hall and Iglica, shows that this method effectively preserves the statistical properties of the radar data, generating high‐fidelity SPMs. This approach significantly improves landmark identification compared to classical correlation methods, paving the way for more robust and resilient UAV navigation systems.
- 1 January 2025
- Published by Institution of Engineering and Technology (IET) in IET Radar, Sonar & Navigation
- Vol. 19 (1)
Open AccessHigh‐Rise Architectural Landmarks Detection and Identification by Spatio‐Probabilistic Models for UAV Anti‐Collision Radar—A Real Data Case
Unmanned aerial vehicles (UAVs) heavily rely on GPS, a system vulnerable to signal interference in complex urban environments. Although radar systems offer a robust alternative due to their ability to effectively penetrate adverse weather and operate in darkness, a key challenge remains: reliably identifying static architectural landmarks from sparse and noisy radar echoes. This paper proposes a novel method for creating spatio‐probabilistic models (SPMs) of radar echoes from high‐rise urban landmarks, enabling independent, radar‐based UAV localisation. We employ kernel density estimation on real radar data, acquired with a custom‐designed X‐band ENAVI radar, focusing on large arena buildings and slender spires. These SPMs are then used to detect and identify landmarks by calculating the divergence between the probability distributions of the real‐time received echoes and the preestimated reference models. Our evaluation, using probabilistic divergence metrics on Wrocław's Centennial Hall and Iglica, shows that this method effectively preserves the statistical properties of the radar data, generating high‐fidelity SPMs. This approach significantly improves landmark identification compared to classical correlation methods, paving the way for more robust and resilient UAV navigation systems.