Pub Date : 2025-07-08DOI: 10.1109/TRS.2025.3586927
Yujie Zhang;Xueru Bai;Feng Zhou
The maneuvering of the targets will induce time-varying Doppler during observation, posing great challenges for well-focused inverse synthetic aperture radar (ISAR) imaging. Furthermore, ISAR may encounter complex observation conditions such as incomplete data and low signal-to-noise ratio (SNR), which render the conventional maneuvering targets imaging methods invalid. To address these issues, this article proposes a novel high-resolution ISAR imaging method of maneuvering targets. First, the sparse imaging model of maneuvering targets is constructed by incorporating the rotation parameters into the observation dictionary. Then, the gamma-complex Gaussian prior is assigned to the ISAR image to exploit its sparse nature. On this basis, to circumvent the matrix inversion embedded in the traditional sparse Bayesian learning (SBL) method, the model lower bound is relaxed and a novel algorithm is proposed for efficient ISAR image reconstruction, dubbed 2-D complex fast SBL (2D-CFSBL). Furthermore, the maximum likelihood estimation is utilized to estimate the rotation parameters accurately. Finally, ISAR image reconstruction and rotation parameters estimation are performed iteratively to obtain well-focused image. Experimental results have validated the effectiveness and superiority of the proposed method under incomplete data and low SNR conditions.
{"title":"High-Resolution ISAR Imaging of Maneuvering Targets Based on 2-D Complex Fast Sparse Bayesian Learning","authors":"Yujie Zhang;Xueru Bai;Feng Zhou","doi":"10.1109/TRS.2025.3586927","DOIUrl":"https://doi.org/10.1109/TRS.2025.3586927","url":null,"abstract":"The maneuvering of the targets will induce time-varying Doppler during observation, posing great challenges for well-focused inverse synthetic aperture radar (ISAR) imaging. Furthermore, ISAR may encounter complex observation conditions such as incomplete data and low signal-to-noise ratio (SNR), which render the conventional maneuvering targets imaging methods invalid. To address these issues, this article proposes a novel high-resolution ISAR imaging method of maneuvering targets. First, the sparse imaging model of maneuvering targets is constructed by incorporating the rotation parameters into the observation dictionary. Then, the gamma-complex Gaussian prior is assigned to the ISAR image to exploit its sparse nature. On this basis, to circumvent the matrix inversion embedded in the traditional sparse Bayesian learning (SBL) method, the model lower bound is relaxed and a novel algorithm is proposed for efficient ISAR image reconstruction, dubbed 2-D complex fast SBL (2D-CFSBL). Furthermore, the maximum likelihood estimation is utilized to estimate the rotation parameters accurately. Finally, ISAR image reconstruction and rotation parameters estimation are performed iteratively to obtain well-focused image. Experimental results have validated the effectiveness and superiority of the proposed method under incomplete data and low SNR conditions.","PeriodicalId":100645,"journal":{"name":"IEEE Transactions on Radar Systems","volume":"3 ","pages":"995-1005"},"PeriodicalIF":0.0,"publicationDate":"2025-07-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144671258","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-07-02DOI: 10.1109/TRS.2025.3584994
Vinay Kulkarni;V. V. Reddy;Neha Maheshwari
This article presents a novel radar signal detection pipeline focused on detecting large targets such as cars and sports utility vehicles (SUVs). Traditional methods, such as ordered-statistic constant false alarm rate (OS-CFAR), commonly used in automotive radar, are designed for point or isotropic target models. These may not adequately capture the range-Doppler (RD) scattering patterns of larger targets, especially in high-resolution radar systems. Additional modules such as association and tracking are necessary to refine and consolidate the detections over multiple dwells. To address these limitations, we propose a detection technique based on the probability density function (pdf) of RD segments, leveraging the Kolmogorov–Arnold neural network (KAN) to learn the data and generate interpretable symbolic expressions for binary hypotheses. Beside the Monte Carlo study showing better performance for the proposed KAN expression over OS-CFAR, it is shown to exhibit a probability of detection ($P_{D}$ ) of 96% when transfer learned with field data. The false alarm rate ($P_{mathrm {FA}}$ ) is comparable with OS-CFAR designed with $P_{mathrm {FA}}=10^{-6}$ . The study also examines how the number of pdf bins in the RD segment affects the performance of KAN-based detection.
{"title":"KAN-Powered Large-Target Detection for Automotive Radar","authors":"Vinay Kulkarni;V. V. Reddy;Neha Maheshwari","doi":"10.1109/TRS.2025.3584994","DOIUrl":"https://doi.org/10.1109/TRS.2025.3584994","url":null,"abstract":"This article presents a novel radar signal detection pipeline focused on detecting large targets such as cars and sports utility vehicles (SUVs). Traditional methods, such as ordered-statistic constant false alarm rate (OS-CFAR), commonly used in automotive radar, are designed for point or isotropic target models. These may not adequately capture the range-Doppler (RD) scattering patterns of larger targets, especially in high-resolution radar systems. Additional modules such as association and tracking are necessary to refine and consolidate the detections over multiple dwells. To address these limitations, we propose a detection technique based on the probability density function (pdf) of RD segments, leveraging the Kolmogorov–Arnold neural network (KAN) to learn the data and generate interpretable symbolic expressions for binary hypotheses. Beside the Monte Carlo study showing better performance for the proposed KAN expression over OS-CFAR, it is shown to exhibit a probability of detection (<inline-formula> <tex-math>$P_{D}$ </tex-math></inline-formula>) of 96% when transfer learned with field data. The false alarm rate (<inline-formula> <tex-math>$P_{mathrm {FA}}$ </tex-math></inline-formula>) is comparable with OS-CFAR designed with <inline-formula> <tex-math>$P_{mathrm {FA}}=10^{-6}$ </tex-math></inline-formula>. The study also examines how the number of pdf bins in the RD segment affects the performance of KAN-based detection.","PeriodicalId":100645,"journal":{"name":"IEEE Transactions on Radar Systems","volume":"3 ","pages":"963-968"},"PeriodicalIF":0.0,"publicationDate":"2025-07-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144597889","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Deep neural networks (DNNs) are extensively employed in synthetic aperture radar (SAR) automatic target recognition (ATR) systems; however, their security and reliability pose significant challenges in this high-risk domain. While considerable efforts have been made to address the vulnerability of DNNs to adversarial attacks, the SAR ATR community has not yet devoted substantial resources to investigating the newly emerging security risks associated with backdoor attacks, which are more threatening because of their attack flexibility, high stealthiness, and versatile attack modes. To investigate backdoor attacks in SAR ATR, we present an innovative method named ASCM-based physical backdoor attack (AMPBA), which generates a physically realizable trigger with clear electromagnetic characteristics and physical attributes based on the attributed scattering center model (ASCM). Specifically, the AMPBA embeds the trigger into limited training samples to produce a poisoned training dataset; after that, training of a DNN-based classifier would inject into it a stealthy backdoor that can be activated by the trigger (either digitally mimicking that of training or physically in practice for real-time attacks). To further enhance the threat level and practicability of the proposed AMPBA, we additionally propose a backdoor attack strategy called low-intensity training and high-intensity inference (LTHI), which utilizes low-intensity triggers during training to maximize stealthiness and high-intensity triggers during inference for enhanced attack performance. Extensive experiments based on the representative MSTAR dataset validate the effectiveness, stealthiness, and robustness of our AMPBA, which, alternatively, highlight the importance of designing effective backdoor defense mechanisms for high-risk applications.
{"title":"Stealthy Backdoor Attack in SAR Target Recognition With ASCM-Based Physically Realizable Triggers","authors":"Fei Zeng;Yuanjia Chen;Yulai Cong;Lei Zhang;Sijia Li;Jianqiang Xu;Jia Duan","doi":"10.1109/TRS.2025.3582438","DOIUrl":"https://doi.org/10.1109/TRS.2025.3582438","url":null,"abstract":"Deep neural networks (DNNs) are extensively employed in synthetic aperture radar (SAR) automatic target recognition (ATR) systems; however, their security and reliability pose significant challenges in this high-risk domain. While considerable efforts have been made to address the vulnerability of DNNs to adversarial attacks, the SAR ATR community has not yet devoted substantial resources to investigating the newly emerging security risks associated with backdoor attacks, which are more threatening because of their attack flexibility, high stealthiness, and versatile attack modes. To investigate backdoor attacks in SAR ATR, we present an innovative method named ASCM-based physical backdoor attack (AMPBA), which generates a physically realizable trigger with clear electromagnetic characteristics and physical attributes based on the attributed scattering center model (ASCM). Specifically, the AMPBA embeds the trigger into limited training samples to produce a poisoned training dataset; after that, training of a DNN-based classifier would inject into it a stealthy backdoor that can be activated by the trigger (either digitally mimicking that of training or physically in practice for real-time attacks). To further enhance the threat level and practicability of the proposed AMPBA, we additionally propose a backdoor attack strategy called low-intensity training and high-intensity inference (LTHI), which utilizes low-intensity triggers during training to maximize stealthiness and high-intensity triggers during inference for enhanced attack performance. Extensive experiments based on the representative MSTAR dataset validate the effectiveness, stealthiness, and robustness of our AMPBA, which, alternatively, highlight the importance of designing effective backdoor defense mechanisms for high-risk applications.","PeriodicalId":100645,"journal":{"name":"IEEE Transactions on Radar Systems","volume":"3 ","pages":"947-962"},"PeriodicalIF":0.0,"publicationDate":"2025-06-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144581546","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-06-20DOI: 10.1109/TRS.2025.3581923
Longfei Wang;Zhunga Liu;Zuowei Zhang;Xiaokui Yue
In the field of remote sensing target recognition, the fusion of synthetic aperture radar (SAR) and optical targets faces significant challenges due to the huge differences in feature representation. Current fusion recognition methods primarily focus on the feature alignment, overlooking the effective utilization of the distinct features inherent to each modality. A decoupled contrastive learning framework constrained by incoherent entropy (DCL-IE) is proposed to fuse the differential features of both SAR and optical modalities. DCL-IE can effectively enhance the model’s ability to distinguish between interclass differences between SAR and optical targets, thereby improving the accuracy of SAR target recognition. Specifically, decoupled contrastive learning (DCL) is designed to efficiently concentrate on different class features when oriented to cross-modal differential representations. The proposed relaxed label assignment algorithm can effectively distinguish between one specific class and the other classes, promoting the extension of DCL into the unsupervised learning domain. Furthermore, the physical incoherent entropy (IE) features are utilized to guide the learning direction of interclass representations, which enhances the extraction of intraclass features by leveraging frequency robustness. Extensive experiments with various target recognition methods on SAR and optical datasets, including FUSAR-Ship, FGSC-23, and FGSCR-42, demonstrate the effectiveness of the proposed framework.
{"title":"Decoupled Contrastive Learning Constrained by Physical Feature for SAR Target Recognition","authors":"Longfei Wang;Zhunga Liu;Zuowei Zhang;Xiaokui Yue","doi":"10.1109/TRS.2025.3581923","DOIUrl":"https://doi.org/10.1109/TRS.2025.3581923","url":null,"abstract":"In the field of remote sensing target recognition, the fusion of synthetic aperture radar (SAR) and optical targets faces significant challenges due to the huge differences in feature representation. Current fusion recognition methods primarily focus on the feature alignment, overlooking the effective utilization of the distinct features inherent to each modality. A decoupled contrastive learning framework constrained by incoherent entropy (DCL-IE) is proposed to fuse the differential features of both SAR and optical modalities. DCL-IE can effectively enhance the model’s ability to distinguish between interclass differences between SAR and optical targets, thereby improving the accuracy of SAR target recognition. Specifically, decoupled contrastive learning (DCL) is designed to efficiently concentrate on different class features when oriented to cross-modal differential representations. The proposed relaxed label assignment algorithm can effectively distinguish between one specific class and the other classes, promoting the extension of DCL into the unsupervised learning domain. Furthermore, the physical incoherent entropy (IE) features are utilized to guide the learning direction of interclass representations, which enhances the extraction of intraclass features by leveraging frequency robustness. Extensive experiments with various target recognition methods on SAR and optical datasets, including FUSAR-Ship, FGSC-23, and FGSCR-42, demonstrate the effectiveness of the proposed framework.","PeriodicalId":100645,"journal":{"name":"IEEE Transactions on Radar Systems","volume":"3 ","pages":"935-946"},"PeriodicalIF":0.0,"publicationDate":"2025-06-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144519459","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Advancement toward fully autonomous systems requires enhanced sensing and perception, particularly a 360° vision for safe maneuvering. One approach to achieving this is through a distributed network of radar sensors, operating in homogeneous or heterogeneous configurations, strategically positioned to provide increased coverage and visibility in otherwise blind regions. Such a multiperspective sensing network, complemented with multimodal signal processing, can significantly improve the angular resolution of the radar, delivering high-fidelity scene imagery essential for region classification and path planning. This study presents a methodology for multimodal and multiperspective sensing using heterogeneous radar sensors, utilizing Doppler beam sharpening (DBS) within multiple-input-multiple-output (MIMO) radars to enhance the resolution and coverage. Traditional frequency-modulated continuous wave (FMCW)–MIMO radars, currently the most widely used configuration, are prone to Doppler aliasing, limiting the field of view (FoV) in DBS and MIMO–DBS processing. To address this limitation, the effective FoV in multiperspective image is extended to that provided by the radar’s physical aperture. The proposed framework is validated using 77-GHz radar chipsets in both automotive and maritime conditions, with sensors mounted in front-looking, corner-looking, and side-looking orientations.
{"title":"High-Resolution Augmented Multimodal Sensing of Distributed Radar Network","authors":"Anum Pirkani;Dillon Kumar;Edward Hoare;Muge Bekar;Natalie Reeves;Mikhail Cherniakov;Marina Gashinova","doi":"10.1109/TRS.2025.3581396","DOIUrl":"https://doi.org/10.1109/TRS.2025.3581396","url":null,"abstract":"Advancement toward fully autonomous systems requires enhanced sensing and perception, particularly a 360° vision for safe maneuvering. One approach to achieving this is through a distributed network of radar sensors, operating in homogeneous or heterogeneous configurations, strategically positioned to provide increased coverage and visibility in otherwise blind regions. Such a multiperspective sensing network, complemented with multimodal signal processing, can significantly improve the angular resolution of the radar, delivering high-fidelity scene imagery essential for region classification and path planning. This study presents a methodology for multimodal and multiperspective sensing using heterogeneous radar sensors, utilizing Doppler beam sharpening (DBS) within multiple-input-multiple-output (MIMO) radars to enhance the resolution and coverage. Traditional frequency-modulated continuous wave (FMCW)–MIMO radars, currently the most widely used configuration, are prone to Doppler aliasing, limiting the field of view (FoV) in DBS and MIMO–DBS processing. To address this limitation, the effective FoV in multiperspective image is extended to that provided by the radar’s physical aperture. The proposed framework is validated using 77-GHz radar chipsets in both automotive and maritime conditions, with sensors mounted in front-looking, corner-looking, and side-looking orientations.","PeriodicalId":100645,"journal":{"name":"IEEE Transactions on Radar Systems","volume":"3 ","pages":"905-918"},"PeriodicalIF":0.0,"publicationDate":"2025-06-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144557960","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
The objective of zero-shot synthetic aperture radar (SAR) image target recognition is to identify the novel unobserved targets for which no training samples are available. The zero-shot recognition method for SAR targets merits investigation, where using electromagnetic simulated images as training data is a viable approach. Nevertheless, the networks trained on the simulated images exhibit difficulty in generalizing to the real images due to the inherent discrepancies in the distribution of the simulated and the real domains. The majority of existing research employs unsupervised domain adaptation methods to address such cross-domain recognition problems. However, these methods are not applicable in zero-shot scenarios, as they require the availability of unlabeled real data from unknown classes during training. Therefore, to address the challenging issue of zero-shot cross-domain recognition for SAR targets, a zero-shot domain adaptation (ZSDA) for SAR target recognition based on cooperative learning of domain alignment and task alignment is proposed. Specifically, we perform domain adaptation using the simulated and real data from the seen classes, to ensure that this alignment can be generalized to the unseen classes. First, a transfer-weighted domain adversarial learning method is proposed to achieve a more robust domain alignment of the seen classes. Second, a classification-based adversarial learning method is proposed to achieve task alignment between the seen and unseen classes within two domains. Finally, a feature fusion refinement module is proposed for the cooperative learning of the two alignment processes. In the context of collaborative learning, task alignment facilitates the transfer of the domain alignment learned from the seen classes to the unseen classes. The experimental results demonstrate the efficacy of the proposed method in SAR zero-shot cross-domain recognition, achieving recognition accuracies of 91.68%, 85.83%, 83.90%, and 77.73% for three unseen class real images across four distinct experimental groups, surpassing the current state-of-the-art methods.
{"title":"Zero-Shot Domain Adaptation for SAR Target Recognition Based on Cooperative Learning of Domain Alignment and Task Alignment","authors":"Guo Chen;Siqian Zhang;Zheng Zhou;Lingjun Zhao;Gangyao Kuang","doi":"10.1109/TRS.2025.3580543","DOIUrl":"https://doi.org/10.1109/TRS.2025.3580543","url":null,"abstract":"The objective of zero-shot synthetic aperture radar (SAR) image target recognition is to identify the novel unobserved targets for which no training samples are available. The zero-shot recognition method for SAR targets merits investigation, where using electromagnetic simulated images as training data is a viable approach. Nevertheless, the networks trained on the simulated images exhibit difficulty in generalizing to the real images due to the inherent discrepancies in the distribution of the simulated and the real domains. The majority of existing research employs unsupervised domain adaptation methods to address such cross-domain recognition problems. However, these methods are not applicable in zero-shot scenarios, as they require the availability of unlabeled real data from unknown classes during training. Therefore, to address the challenging issue of zero-shot cross-domain recognition for SAR targets, a zero-shot domain adaptation (ZSDA) for SAR target recognition based on cooperative learning of domain alignment and task alignment is proposed. Specifically, we perform domain adaptation using the simulated and real data from the seen classes, to ensure that this alignment can be generalized to the unseen classes. First, a transfer-weighted domain adversarial learning method is proposed to achieve a more robust domain alignment of the seen classes. Second, a classification-based adversarial learning method is proposed to achieve task alignment between the seen and unseen classes within two domains. Finally, a feature fusion refinement module is proposed for the cooperative learning of the two alignment processes. In the context of collaborative learning, task alignment facilitates the transfer of the domain alignment learned from the seen classes to the unseen classes. The experimental results demonstrate the efficacy of the proposed method in SAR zero-shot cross-domain recognition, achieving recognition accuracies of 91.68%, 85.83%, 83.90%, and 77.73% for three unseen class real images across four distinct experimental groups, surpassing the current state-of-the-art methods.","PeriodicalId":100645,"journal":{"name":"IEEE Transactions on Radar Systems","volume":"3 ","pages":"890-904"},"PeriodicalIF":0.0,"publicationDate":"2025-06-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144557959","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-06-17DOI: 10.1109/TRS.2025.3580623
Hajar Abedi;Jenna Hall;Ji Beom Bae;Plinio P. Morita;Alexander Wong;Jennifer Boger;George Shaker
Gait analysis is one of the most useful predictors of disease in older adults, but it is not always practical for physicians to monitor. This article aimed to create a system that could continuously and reliably monitor gait patterns of varying step lengths and speeds in cluttered environments, enabling around-the-clock monitoring in personal living spaces. This novel study uses multiple input multiple output frequency-modulated continuous-wave (MIMO FMCW) radar to track nonlinear movement in cluttered environments designed to replicate a living space in a home. A subjects tracker and association (STA) algorithm was proposed to distinguish direct signals with multipath effects and remove ghost signals created by clutter. Six participants were instructed to walk along designated paths with varied step lengths (30, 60, and 80 cm), and our findings supported the system’s ability to capture walking speed, step count, and step length. The system was successful in accurately tracking gait parameters in naturalistic settings, offering a potential solution to autonomous, continuous in-home gait analysis.
{"title":"Continuous In-Home Gait Analysis Using FMCW Radar in Naturalistic Environments","authors":"Hajar Abedi;Jenna Hall;Ji Beom Bae;Plinio P. Morita;Alexander Wong;Jennifer Boger;George Shaker","doi":"10.1109/TRS.2025.3580623","DOIUrl":"https://doi.org/10.1109/TRS.2025.3580623","url":null,"abstract":"Gait analysis is one of the most useful predictors of disease in older adults, but it is not always practical for physicians to monitor. This article aimed to create a system that could continuously and reliably monitor gait patterns of varying step lengths and speeds in cluttered environments, enabling around-the-clock monitoring in personal living spaces. This novel study uses multiple input multiple output frequency-modulated continuous-wave (MIMO FMCW) radar to track nonlinear movement in cluttered environments designed to replicate a living space in a home. A subjects tracker and association (STA) algorithm was proposed to distinguish direct signals with multipath effects and remove ghost signals created by clutter. Six participants were instructed to walk along designated paths with varied step lengths (30, 60, and 80 cm), and our findings supported the system’s ability to capture walking speed, step count, and step length. The system was successful in accurately tracking gait parameters in naturalistic settings, offering a potential solution to autonomous, continuous in-home gait analysis.","PeriodicalId":100645,"journal":{"name":"IEEE Transactions on Radar Systems","volume":"3 ","pages":"969-981"},"PeriodicalIF":0.0,"publicationDate":"2025-06-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144597894","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-06-17DOI: 10.1109/TRS.2025.3580606
Xiang Wang;Yumiao Wang;Guolong Cui
Maritime radar detectors developed using deep learning technology have demonstrated promising performance in the clutter environment. However, real clutter environments are usually time-varying, and the nonstationary radar data stream easily breaks the independent and identically distributed (i.i.d.) prerequisite of standard deep learning detectors, decreasing the detector’s performance. This article considers the problem of adaptive maritime radar target detection for deep learning-based detectors in time-varying clutter environments. We propose an adaptive target detection framework based on an active self-learning (SL) strategy, which can actively sense the environment shift and update the detector parameters correspondingly through SL. Specifically, we first use the annotated dataset to train an initial detector. Then, we design an environment sensing module by adding a subdetection head on the detector. When the detector works in time-varying clutter environments, the entropy between the detector’s output and the subdetection head’s output is utilized to sense the environment shift. Next, we propose an SL strategy that combines adaptive pseudo-label generation with consistency regularization. Once the environment shift is detected, the detector parameters are updated by the proposed SL strategy, improving the detector’s performance in time-varying clutter environments. Experimental results on the public maritime radar database validate the effectiveness of the proposed framework.
{"title":"Adaptive Intelligent Radar Target Detection in Time-Varying Sea Clutter via Activate Self-Learning","authors":"Xiang Wang;Yumiao Wang;Guolong Cui","doi":"10.1109/TRS.2025.3580606","DOIUrl":"https://doi.org/10.1109/TRS.2025.3580606","url":null,"abstract":"Maritime radar detectors developed using deep learning technology have demonstrated promising performance in the clutter environment. However, real clutter environments are usually time-varying, and the nonstationary radar data stream easily breaks the independent and identically distributed (i.i.d.) prerequisite of standard deep learning detectors, decreasing the detector’s performance. This article considers the problem of adaptive maritime radar target detection for deep learning-based detectors in time-varying clutter environments. We propose an adaptive target detection framework based on an active self-learning (SL) strategy, which can actively sense the environment shift and update the detector parameters correspondingly through SL. Specifically, we first use the annotated dataset to train an initial detector. Then, we design an environment sensing module by adding a subdetection head on the detector. When the detector works in time-varying clutter environments, the entropy between the detector’s output and the subdetection head’s output is utilized to sense the environment shift. Next, we propose an SL strategy that combines adaptive pseudo-label generation with consistency regularization. Once the environment shift is detected, the detector parameters are updated by the proposed SL strategy, improving the detector’s performance in time-varying clutter environments. Experimental results on the public maritime radar database validate the effectiveness of the proposed framework.","PeriodicalId":100645,"journal":{"name":"IEEE Transactions on Radar Systems","volume":"3 ","pages":"919-934"},"PeriodicalIF":0.0,"publicationDate":"2025-06-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144519460","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
At the current state of the scientific discourse on quantum radar, the best understood and experimentally feasible types of implementation are based on two-mode-squeezed-vacuum (TMSV) photon states and aimed at the task of target detection. The operating environment, in which an advantage over classical radar may be attainable, is therefore limited to the extreme regimes of very low signal-to-noise ratios (SNRs) and high thermal noise levels as well as confining the required hardware at mK temperatures. In this work, we approach the open question of how to optimally operate a potential quantum radar system. To this end, we define the optimal operation using the detection advantage against classical radar as well as the efficient usage of the resource measurement time. We show that there is a tradeoff between time efficiency and outperformance of classical radar and specify the conditions for such an operation. Building on this aspect, we investigate the concept of the fair classical comparison to facilitate the understanding of its relation to quantum radar.
{"title":"A Note on the Efficient Operation of Quantum Radar and the Fair Classical Comparison","authors":"Florian Bischeltsrieder;Michael Würth;Markus Peichl;Wolfgang Utschick","doi":"10.1109/TRS.2025.3579042","DOIUrl":"https://doi.org/10.1109/TRS.2025.3579042","url":null,"abstract":"At the current state of the scientific discourse on quantum radar, the best understood and experimentally feasible types of implementation are based on two-mode-squeezed-vacuum (TMSV) photon states and aimed at the task of target detection. The operating environment, in which an advantage over classical radar may be attainable, is therefore limited to the extreme regimes of very low signal-to-noise ratios (SNRs) and high thermal noise levels as well as confining the required hardware at mK temperatures. In this work, we approach the open question of how to optimally operate a potential quantum radar system. To this end, we define the optimal operation using the detection advantage against classical radar as well as the efficient usage of the resource measurement time. We show that there is a tradeoff between time efficiency and outperformance of classical radar and specify the conditions for such an operation. Building on this aspect, we investigate the concept of the fair classical comparison to facilitate the understanding of its relation to quantum radar.","PeriodicalId":100645,"journal":{"name":"IEEE Transactions on Radar Systems","volume":"3 ","pages":"875-880"},"PeriodicalIF":0.0,"publicationDate":"2025-06-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11032128","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144481894","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-06-12DOI: 10.1109/TRS.2025.3579026
Marc Reinecke;Theresa Noegel;Oliver Sura;Marcel Hoffmann;Peter Gulden;Martin Vossiek
Automotive forward-looking synthetic aperture radar (FL-SAR) has recently attracted research attention, not only for the resolution gain but also for the exceptional signal-to-clutter ratios (SCRs) that can be achieved. However, when utilizing the backprojection (BP) algorithm for FL-SAR, a mirror-target problem emerges, which is attributable to an inherent flaw of image reconstruction with 2-D spatial sampling grids, such as the ones created in FL-SAR. Constructive superposition of ambiguous subapertures produces magnitudes, which can be significantly higher than those of real targets. This causes false detections and severely impacts higher level tasks such as trajectory planning. This article aims to describe the phenomenon of mirror targets using the well-known example of the BP algorithm. Based on a thorough understanding of the undesirable artifacts, four suppression methods to mitigate false detections were developed. Their viability was ensured through simulative tests. Experimental evaluation in real-world measurement scenarios proved the effectiveness and robustness of all methods. A phase coherency-based classification approach yielded the most accurate results by detecting mirror-target-specific features in the images, thereby enhancing FL-SAR’s imaging capabilities.
{"title":"Mitigation of Mirror Targets in Automotive Forward-Looking Synthetic Aperture Radar","authors":"Marc Reinecke;Theresa Noegel;Oliver Sura;Marcel Hoffmann;Peter Gulden;Martin Vossiek","doi":"10.1109/TRS.2025.3579026","DOIUrl":"https://doi.org/10.1109/TRS.2025.3579026","url":null,"abstract":"Automotive forward-looking synthetic aperture radar (FL-SAR) has recently attracted research attention, not only for the resolution gain but also for the exceptional signal-to-clutter ratios (SCRs) that can be achieved. However, when utilizing the backprojection (BP) algorithm for FL-SAR, a mirror-target problem emerges, which is attributable to an inherent flaw of image reconstruction with 2-D spatial sampling grids, such as the ones created in FL-SAR. Constructive superposition of ambiguous subapertures produces magnitudes, which can be significantly higher than those of real targets. This causes false detections and severely impacts higher level tasks such as trajectory planning. This article aims to describe the phenomenon of mirror targets using the well-known example of the BP algorithm. Based on a thorough understanding of the undesirable artifacts, four suppression methods to mitigate false detections were developed. Their viability was ensured through simulative tests. Experimental evaluation in real-world measurement scenarios proved the effectiveness and robustness of all methods. A phase coherency-based classification approach yielded the most accurate results by detecting mirror-target-specific features in the images, thereby enhancing FL-SAR’s imaging capabilities.","PeriodicalId":100645,"journal":{"name":"IEEE Transactions on Radar Systems","volume":"3 ","pages":"982-994"},"PeriodicalIF":0.0,"publicationDate":"2025-06-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144671259","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：