Pub Date : 2023-05-01DOI: 10.1109/RadarConf2351548.2023.10149551
Yihan Su, Guangbin Zhang, Tianyao Huang, Yimin Liu, Xiqin Wang
Direction finding in partly calibrated arrays, a distributed array with errors between subarrays, receives wide studies. Recently, sparse recovery is used to exploit the blockand rank- sparsity of the signals to self-calibrate the errors and recover the directions, which achieves good performance. Compared with traditional methods based on subspace separation, sparse recovery methods are less sensitive to few snapshots and correlated sources. However, existing sparse recovery methods solve a complex semi-definite programming (SDP) problem, which suffers from high time and space complexity. To this end, we consider to introduce sparse Bayesian learning (SBL) to partly calibrated arrays instead. In a SBL framework, we formulate a sparse recovery problem with self-calibration on errors, and derive the closed-form iterations to solve the problem. Simulations show the feasibility of our proposed method and less time complexity than existing sparse recovery methods.
{"title":"Direction Finding in Partly Calibrated Arrays Using Sparse Bayesian Learning","authors":"Yihan Su, Guangbin Zhang, Tianyao Huang, Yimin Liu, Xiqin Wang","doi":"10.1109/RadarConf2351548.2023.10149551","DOIUrl":"https://doi.org/10.1109/RadarConf2351548.2023.10149551","url":null,"abstract":"Direction finding in partly calibrated arrays, a distributed array with errors between subarrays, receives wide studies. Recently, sparse recovery is used to exploit the blockand rank- sparsity of the signals to self-calibrate the errors and recover the directions, which achieves good performance. Compared with traditional methods based on subspace separation, sparse recovery methods are less sensitive to few snapshots and correlated sources. However, existing sparse recovery methods solve a complex semi-definite programming (SDP) problem, which suffers from high time and space complexity. To this end, we consider to introduce sparse Bayesian learning (SBL) to partly calibrated arrays instead. In a SBL framework, we formulate a sparse recovery problem with self-calibration on errors, and derive the closed-form iterations to solve the problem. Simulations show the feasibility of our proposed method and less time complexity than existing sparse recovery methods.","PeriodicalId":168311,"journal":{"name":"2023 IEEE Radar Conference (RadarConf23)","volume":"23 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121219556","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 : 2023-05-01DOI: 10.1109/RadarConf2351548.2023.10149570
L. Mailaender, A. Lackpour
Dynamic spectrum sharing between airborne radars and 5G cellular networks has the potential for granting additional RF spectrum to cellular networks while preserving the performance of airborne radars. In the case of an airborne radar with a predictably rotating antenna, a spectrum sharing controller can use estimates of the radar's location and beam orientation to anticipate and mitigate RF interference events over a large geographic area. However, localization of the radar is complicated by airborne radar's relatively narrow beamwidth and time-varying waveform. We introduce the Rotating Beam Time-of-Arrival (RB-TOA) algorithm to jointly estimate the radar's location and antenna main beam orientation. Each RF sensor is coarsely time-synchronized and measures the peak of the received signal envelope over each rotation interval to estimate when the radar's main beam maximally couples with the sensor's antenna; these time estimates are then combined at a sensor fusion server and the radar's main beam orientation and location are jointly solved using a gradient descent algorithm. We show that the RBTOA algorithm rapidly converges to a geolocation accuracy that is 50x better than the performance of a two-antenna angle-of-arrival algorithm (AoA) for the same number of sensors.
{"title":"Time-Based Geolocation and Main Beam Estimation of an Airborne Rotating Radar for Spectrum Sharing","authors":"L. Mailaender, A. Lackpour","doi":"10.1109/RadarConf2351548.2023.10149570","DOIUrl":"https://doi.org/10.1109/RadarConf2351548.2023.10149570","url":null,"abstract":"Dynamic spectrum sharing between airborne radars and 5G cellular networks has the potential for granting additional RF spectrum to cellular networks while preserving the performance of airborne radars. In the case of an airborne radar with a predictably rotating antenna, a spectrum sharing controller can use estimates of the radar's location and beam orientation to anticipate and mitigate RF interference events over a large geographic area. However, localization of the radar is complicated by airborne radar's relatively narrow beamwidth and time-varying waveform. We introduce the Rotating Beam Time-of-Arrival (RB-TOA) algorithm to jointly estimate the radar's location and antenna main beam orientation. Each RF sensor is coarsely time-synchronized and measures the peak of the received signal envelope over each rotation interval to estimate when the radar's main beam maximally couples with the sensor's antenna; these time estimates are then combined at a sensor fusion server and the radar's main beam orientation and location are jointly solved using a gradient descent algorithm. We show that the RBTOA algorithm rapidly converges to a geolocation accuracy that is 50x better than the performance of a two-antenna angle-of-arrival algorithm (AoA) for the same number of sensors.","PeriodicalId":168311,"journal":{"name":"2023 IEEE Radar Conference (RadarConf23)","volume":"13 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121846550","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 : 2023-05-01DOI: 10.1109/RadarConf2351548.2023.10149721
K. Jędrzejewski, M. Malanowski, K. Kulpa, M. Pożoga, A. Modrzewski, Michal Karwacki
The paper presents the concept and results of experiments devoted to verifying empirically the potential capability of long-distance space object observation by radar system employing an astronomical LOFAR (LOw-Frequency ARray) radio telescope and a non-cooperative radar illuminator operating in a VHF band. The large antenna array of one of the LOFAR radio telescopes was used as a surveillance receiver to collect weak echo signals reflected from space objects, while the reference signal was recorded by a simple software-defined radio receiver located near the radar illuminator. A dedicated object motion compensation procedure has been applied to detect high-speed space targets in low-Earth orbit. The results of the conducted experiments confirm the possibility of detecting space objects employing the antenna arrays used in the LOFAR radio telescopes and signals emitted by non-cooperative radars to illuminate space objects.
{"title":"Long-Distance Bistatic Measurements of Space Object Motion using LOFAR Radio Telescope and Non-cooperative Radar Illuminator","authors":"K. Jędrzejewski, M. Malanowski, K. Kulpa, M. Pożoga, A. Modrzewski, Michal Karwacki","doi":"10.1109/RadarConf2351548.2023.10149721","DOIUrl":"https://doi.org/10.1109/RadarConf2351548.2023.10149721","url":null,"abstract":"The paper presents the concept and results of experiments devoted to verifying empirically the potential capability of long-distance space object observation by radar system employing an astronomical LOFAR (LOw-Frequency ARray) radio telescope and a non-cooperative radar illuminator operating in a VHF band. The large antenna array of one of the LOFAR radio telescopes was used as a surveillance receiver to collect weak echo signals reflected from space objects, while the reference signal was recorded by a simple software-defined radio receiver located near the radar illuminator. A dedicated object motion compensation procedure has been applied to detect high-speed space targets in low-Earth orbit. The results of the conducted experiments confirm the possibility of detecting space objects employing the antenna arrays used in the LOFAR radio telescopes and signals emitted by non-cooperative radars to illuminate space objects.","PeriodicalId":168311,"journal":{"name":"2023 IEEE Radar Conference (RadarConf23)","volume":"44 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115794224","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 : 2023-05-01DOI: 10.1109/RadarConf2351548.2023.10149729
K. Fatseas, M. Bekooij
Automotive radar and camera fusion relies on linear point transformations from one sensor's coordinate system to the other. However, these transformations cannot handle non-linear dynamics and are susceptible to sensor noise. Furthermore, they operate on a point-to-point basis, so it is impossible to capture all the characteristics of an object. This paper introduces a method that performs detection-to-detection association by projecting heterogeneous object features from the two sensors into a common high-dimensional space. We associate 2D bounding boxes and radar detections based on the Euclidean distance between their projections. Our method utilizes deep neural networks to transform feature vectors instead of single points. Therefore, we can leverage real-world data to learn non-linear dynamics and utilize several features to provide a better description for each object. We evaluate our association method against a traditional rule-based method, showing that it improves the accuracy of the association algorithm and it is more robust in complex scenarios with multiple objects.
{"title":"Association of Camera and Radar Detections Using Neural Networks","authors":"K. Fatseas, M. Bekooij","doi":"10.1109/RadarConf2351548.2023.10149729","DOIUrl":"https://doi.org/10.1109/RadarConf2351548.2023.10149729","url":null,"abstract":"Automotive radar and camera fusion relies on linear point transformations from one sensor's coordinate system to the other. However, these transformations cannot handle non-linear dynamics and are susceptible to sensor noise. Furthermore, they operate on a point-to-point basis, so it is impossible to capture all the characteristics of an object. This paper introduces a method that performs detection-to-detection association by projecting heterogeneous object features from the two sensors into a common high-dimensional space. We associate 2D bounding boxes and radar detections based on the Euclidean distance between their projections. Our method utilizes deep neural networks to transform feature vectors instead of single points. Therefore, we can leverage real-world data to learn non-linear dynamics and utilize several features to provide a better description for each object. We evaluate our association method against a traditional rule-based method, showing that it improves the accuracy of the association algorithm and it is more robust in complex scenarios with multiple objects.","PeriodicalId":168311,"journal":{"name":"2023 IEEE Radar Conference (RadarConf23)","volume":"397 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131963978","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 : 2023-05-01DOI: 10.1109/RadarConf2351548.2023.10149580
R. Blázquez-García, D. Cristallini, M. Ummenhofer, V. Seidel, J. Heckenbach, D. O’Hagan
Given the limited available information about Star-link and OneWeb signals but the relevant capabilities that they may provide, this work is focused on the experimental acquisition and comparison for passive radar applications of the user downlink signals transmitted by these emerging satellite constellations. In order to received these signal, an updated version of the SABBIA system has been developed with satellite tracking capabilities and enhanced instantaneous bandwidth to enable the digitization of a complete transmission channel. Based on the analysis of the received Starlink and OneWeb signals in terms of the ambiguity function, both constellations are considered suitable as complementary potential illuminations of opportunity for high-resolution passive radar applications.
{"title":"Experimental comparison of Starlink and OneWeb signals for passive radar","authors":"R. Blázquez-García, D. Cristallini, M. Ummenhofer, V. Seidel, J. Heckenbach, D. O’Hagan","doi":"10.1109/RadarConf2351548.2023.10149580","DOIUrl":"https://doi.org/10.1109/RadarConf2351548.2023.10149580","url":null,"abstract":"Given the limited available information about Star-link and OneWeb signals but the relevant capabilities that they may provide, this work is focused on the experimental acquisition and comparison for passive radar applications of the user downlink signals transmitted by these emerging satellite constellations. In order to received these signal, an updated version of the SABBIA system has been developed with satellite tracking capabilities and enhanced instantaneous bandwidth to enable the digitization of a complete transmission channel. Based on the analysis of the received Starlink and OneWeb signals in terms of the ambiguity function, both constellations are considered suitable as complementary potential illuminations of opportunity for high-resolution passive radar applications.","PeriodicalId":168311,"journal":{"name":"2023 IEEE Radar Conference (RadarConf23)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130446437","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}
Millimeter-wave (mmW) automotive radar imaging technology is widely applied to advanced driver assistance systems (ADAS). Existing super-resolution imaging methods can improve angular resolution for automotive radar with a limited aperture. However, these super-resolution methods have high computational complexity, meanwhile have poor imaging performance in single-snapshot. To address these problems. we propose a fast 2D super-resolution imaging method for real-time and high-quality automotive radar imaging. First, a novel Bayesian compressive sensing with the Kailath-Variant (BCS-KV) imaging method is proposed to achieve superior angular super-resolution in single-snapshot. And the K-V is used to reduce the complexity of matrix inversion. Then, in the range dimension, a Multi-Channel Accumulation (MCA) is utilized to detect the effective range unit to further reduce the 2D imaging computational complexity. Finally, both simulated and experimental results demonstrate that the proposed method has lower computational complexity and compelling imaging performance than other imaging methods.
{"title":"A Fast 2D Super-resolution Imaging Method via Bayesian Compressive Sensing for mmWave Automotive radar","authors":"Yanqin Xu, Yuan Song, Shunjun Wei, Xiaoling Zhang, Lanwei Guo, Xiaowo Xu","doi":"10.1109/RadarConf2351548.2023.10149712","DOIUrl":"https://doi.org/10.1109/RadarConf2351548.2023.10149712","url":null,"abstract":"Millimeter-wave (mmW) automotive radar imaging technology is widely applied to advanced driver assistance systems (ADAS). Existing super-resolution imaging methods can improve angular resolution for automotive radar with a limited aperture. However, these super-resolution methods have high computational complexity, meanwhile have poor imaging performance in single-snapshot. To address these problems. we propose a fast 2D super-resolution imaging method for real-time and high-quality automotive radar imaging. First, a novel Bayesian compressive sensing with the Kailath-Variant (BCS-KV) imaging method is proposed to achieve superior angular super-resolution in single-snapshot. And the K-V is used to reduce the complexity of matrix inversion. Then, in the range dimension, a Multi-Channel Accumulation (MCA) is utilized to detect the effective range unit to further reduce the 2D imaging computational complexity. Finally, both simulated and experimental results demonstrate that the proposed method has lower computational complexity and compelling imaging performance than other imaging methods.","PeriodicalId":168311,"journal":{"name":"2023 IEEE Radar Conference (RadarConf23)","volume":"36 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134143017","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 : 2023-05-01DOI: 10.1109/RadarConf2351548.2023.10149563
Jiaheng Wang, Yalong Wang, Haoqi Wu, Zhihang Wang, Jun Yu Li
With the Kronecker product structure constraint, this paper proposes a covariance matrix (CM) estimation method in the Compound-Gaussian (CG) sea clutter background. We assume the CG clutter in different polarization channels has different textures, which is different from the existing Kronecker structure-based CM estimation methods for polarimetric target detection. Based on the maximum likelihood (ML) criterion, we obtain the fixed point equation of the CM and solve it by an iterative algorithm. The proposed method is referred to as the Kronecker-based maximum likelihood estimate (KMLE), and the relevance of KMLE to the existing estimation methods is also discussed. For the performance assessment, we demonstrate the estimation accuracy of KMLE by presenting the normalized mean-square error (NMSE), and the detection performance is assessed by inserting the estimated CM into the test statistic of the texture-free generalized likelihood ratio test (TF-GLRT) detector. Through simulations with the synthetic and real sea clutter, we verify that KMLE outperforms other estimation methods when the training samples are limited.
{"title":"Covariance Matrix Estimation With Kronecker Structure Constraint For Polarimetric Detection","authors":"Jiaheng Wang, Yalong Wang, Haoqi Wu, Zhihang Wang, Jun Yu Li","doi":"10.1109/RadarConf2351548.2023.10149563","DOIUrl":"https://doi.org/10.1109/RadarConf2351548.2023.10149563","url":null,"abstract":"With the Kronecker product structure constraint, this paper proposes a covariance matrix (CM) estimation method in the Compound-Gaussian (CG) sea clutter background. We assume the CG clutter in different polarization channels has different textures, which is different from the existing Kronecker structure-based CM estimation methods for polarimetric target detection. Based on the maximum likelihood (ML) criterion, we obtain the fixed point equation of the CM and solve it by an iterative algorithm. The proposed method is referred to as the Kronecker-based maximum likelihood estimate (KMLE), and the relevance of KMLE to the existing estimation methods is also discussed. For the performance assessment, we demonstrate the estimation accuracy of KMLE by presenting the normalized mean-square error (NMSE), and the detection performance is assessed by inserting the estimated CM into the test statistic of the texture-free generalized likelihood ratio test (TF-GLRT) detector. Through simulations with the synthetic and real sea clutter, we verify that KMLE outperforms other estimation methods when the training samples are limited.","PeriodicalId":168311,"journal":{"name":"2023 IEEE Radar Conference (RadarConf23)","volume":"178 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134362225","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 : 2023-05-01DOI: 10.1109/RadarConf2351548.2023.10149705
Johannes Benz, Christian Weiss, Axel Acosta Aponte, Gor Hakobyan
Development of machine learning (ML) models requires large amounts of labeled data. For safety critical automotive applications such as radar based perception, the dataset must contain various and rare corner cases, e.g. rare instances that have not been seen before. The straightforward approach of measuring and manually labeling large amounts of data to capture such corner cases is often infeasible or impractical. Thus, approaches for efficiently selecting and labeling the relevant data are essential for ML-based radar applications. In this paper, we propose a method for semi-supervised learning (SSL) for radar object type classification. We use the track consistency of tracked radar objects as a constraint to generate high-quality labels for the vast portions of the unlabeled dataset. We extend the proposed SSL approach with active learning that considers the data relevance, such that the most relevant data with the least accurate auto-labels are selected for human labeling. We show that the proposed approach achieves a saving of more than 87% of human labeling costs based on auto-labeling and relevant data selection.
{"title":"Semi-Supervised Active Learning for Radar based Object Classification Using Track Consistency","authors":"Johannes Benz, Christian Weiss, Axel Acosta Aponte, Gor Hakobyan","doi":"10.1109/RadarConf2351548.2023.10149705","DOIUrl":"https://doi.org/10.1109/RadarConf2351548.2023.10149705","url":null,"abstract":"Development of machine learning (ML) models requires large amounts of labeled data. For safety critical automotive applications such as radar based perception, the dataset must contain various and rare corner cases, e.g. rare instances that have not been seen before. The straightforward approach of measuring and manually labeling large amounts of data to capture such corner cases is often infeasible or impractical. Thus, approaches for efficiently selecting and labeling the relevant data are essential for ML-based radar applications. In this paper, we propose a method for semi-supervised learning (SSL) for radar object type classification. We use the track consistency of tracked radar objects as a constraint to generate high-quality labels for the vast portions of the unlabeled dataset. We extend the proposed SSL approach with active learning that considers the data relevance, such that the most relevant data with the least accurate auto-labels are selected for human labeling. We show that the proposed approach achieves a saving of more than 87% of human labeling costs based on auto-labeling and relevant data selection.","PeriodicalId":168311,"journal":{"name":"2023 IEEE Radar Conference (RadarConf23)","volume":"35 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134407817","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 : 2023-05-01DOI: 10.1109/RadarConf2351548.2023.10149572
Taylan Denizcan Çaha, L. D. Ata
In this study, cognitive radar (CR) applications including radar waveform parameters and track update interval selection are investigated in order to balance the time resource cost and increase the accuracy performance of multiple target tracking systems. For the target tracking part, the unscented Kalman filter (UKF) is applied together with the joint probabilistic data association (JPDA) and the interacting multiple models (IMM) algorithm, which is used to realize more than one target motion model. The waveform parameters and track update interval are adaptively updated by using the outputs of the radar data processing block including target tracking and classification algorithms. The waveform parameters to be updated, the product of the pulse width and the number of integrated pulses, and the track update interval are selected. In the optimization function, the limit values of the parameter selections are decided by using target class information which is supplied by a random forest classifier. Along with the proposed cost function, track continuity and time resource allocation are tested and system performance is demonstrated depending on the target characteristics. In the simulations part, multiple target scenarios that include targets with different maneuvers and radar cross sections (RCS) have been examined and it is shown that the proposed cost function can be applied in multiple target tracking scenarios.
{"title":"Optimization of Waveform Parameters for Multiple Target Tracking Systems in Cognitive Radars","authors":"Taylan Denizcan Çaha, L. D. Ata","doi":"10.1109/RadarConf2351548.2023.10149572","DOIUrl":"https://doi.org/10.1109/RadarConf2351548.2023.10149572","url":null,"abstract":"In this study, cognitive radar (CR) applications including radar waveform parameters and track update interval selection are investigated in order to balance the time resource cost and increase the accuracy performance of multiple target tracking systems. For the target tracking part, the unscented Kalman filter (UKF) is applied together with the joint probabilistic data association (JPDA) and the interacting multiple models (IMM) algorithm, which is used to realize more than one target motion model. The waveform parameters and track update interval are adaptively updated by using the outputs of the radar data processing block including target tracking and classification algorithms. The waveform parameters to be updated, the product of the pulse width and the number of integrated pulses, and the track update interval are selected. In the optimization function, the limit values of the parameter selections are decided by using target class information which is supplied by a random forest classifier. Along with the proposed cost function, track continuity and time resource allocation are tested and system performance is demonstrated depending on the target characteristics. In the simulations part, multiple target scenarios that include targets with different maneuvers and radar cross sections (RCS) have been examined and it is shown that the proposed cost function can be applied in multiple target tracking scenarios.","PeriodicalId":168311,"journal":{"name":"2023 IEEE Radar Conference (RadarConf23)","volume":"67 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131636745","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 : 2023-05-01DOI: 10.1109/RadarConf2351548.2023.10149788
Neeraj Pandey, S. S. Ram
Explainable decision-making is a key component for compliance with regulatory frameworks and winning trust among end users. In this work, we propose to understand the mis-classification of automotive radar images through counterfactual explanations obtained from generative adversarial networks. The proposed method enables perturbations of original radar images belonging to a query class to result in counterfactual images that are classified as the distractor class. The key requirement is that the perturbations must result in realistic images that belong to the original distribution of the query class and also provide physics-based insights into the causes of the misclassification. We test the methods on simulated automotive inverse synthetic aperture radar data images for a query class of a four-wheel mid-size car and a distractor class of a three-wheel auto-rickshaw. Our results show that the shadowing of one or more wheels of the query class is most likely to result in misclassification.
{"title":"Explainable Artificial Intelligence based Classification of Automotive Radar Targets","authors":"Neeraj Pandey, S. S. Ram","doi":"10.1109/RadarConf2351548.2023.10149788","DOIUrl":"https://doi.org/10.1109/RadarConf2351548.2023.10149788","url":null,"abstract":"Explainable decision-making is a key component for compliance with regulatory frameworks and winning trust among end users. In this work, we propose to understand the mis-classification of automotive radar images through counterfactual explanations obtained from generative adversarial networks. The proposed method enables perturbations of original radar images belonging to a query class to result in counterfactual images that are classified as the distractor class. The key requirement is that the perturbations must result in realistic images that belong to the original distribution of the query class and also provide physics-based insights into the causes of the misclassification. We test the methods on simulated automotive inverse synthetic aperture radar data images for a query class of a four-wheel mid-size car and a distractor class of a three-wheel auto-rickshaw. Our results show that the shadowing of one or more wheels of the query class is most likely to result in misclassification.","PeriodicalId":168311,"journal":{"name":"2023 IEEE Radar Conference (RadarConf23)","volume":"225 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130784177","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：