Compressive sensing (CS) theory provides a positive contribution to ISAR imaging. However, the imaging performance of Compressive Sensing Inverse Synthetic Aperture Radar (CS ISAR) imaging methods is limited by the sparsity of the target scene. Dictionary learning (DicL) has been incorporated into CS ISAR imaging better to sparsify the target scene in a certain domain and improve imaging performance. However, the existing dictionary learning-based CS ISAR imaging methods are not adaptive enough and time-consuming. The algorithm parameters need to be manually adjusted for different targets. Their common iterative structure leads to relatively low computational efficiency. To improve the adaptive ability and computational efficiency of DicL-based CS ISAR imaging, we propose an enhanced convolutional DicL-based CS ISAR imaging method. Apart from exploiting the strong learning ability of the multi-layer network structure offered by the convolutional DicL, an attention mapping inferred in both spatial and channel dimensions and a multi-branch convolution are incorporated to enhance the sparsity in the latent space of the Convolutional DicL in CS ISAR imaging. The quantitative and qualitative analyses of the experimental results show that the proposed CS ISAR imaging method outperforms the existing DicL-based CS ISAR imaging methods and is also superior to the typical model-driven DL-based methods like ADMM-net.

In Video synthetic aperture radar (Video SAR) system, the moving target will leave a shadow at its actual position due to Doppler effect. As the shadow of the moving target moves between Video SAR frames, the amplitudes of pixel points at the corresponding positions will jump between frames as well. According to this characteristic, a new method of Video SAR ground moving target detection and tracking based on the inter-frame amplitude temporal curves is proposed in this paper. In this method, the specially designed multiple receptive field fusion neural network model based on frame variation (MRFN-FV) is used to classify the pixel points with obvious inter-frame amplitude jumps on the whole-time axis, and then the false alarms are suppressed based on the temporal change characteristics of pixel points. Finally, the improved clustering algorithm is used to detect, locate and track the moving targets in each frame of SAR images. The effectiveness of the proposed method is verified through the measured data recorded by the THz band Video SAR system.

In this article, a novel method for the moving target detection through multilevel spatial and channelwise attention network based on shallow feature channel module (MSCA-SFCM) is presented, and the circular spotlight (CSL) video synthetic aperture radar ground moving target indication (Video-SAR-GMTI) mode of the Nanjing University of Aeronautics and Astronautics miniature SAR (NUAA MiniSAR) system is introduced. However, due to the lack of moving target samples, MSCA-SFCM cannot be directly applied to the CSL Video-SAR-GMTI mode in the real system. To this end, this article proposes a training sample library construction scheme for moving targets of high verisimilitude. In this scheme, based on the radar system parameters, after the traversal of moving target parameters and SAR imaging, the scattering line characteristic of all possible moving targets under the current system parameters is simulated and then used for MSCA-SFCM network training. Afterward, the properly trained network can be used for moving target detection in real radar data. The effectiveness of the proposed method is verified by the NUAA MiniSAR system.

To improve the classification performance of the convolutional neural network (CNN) for polarimetric synthetic aperture radar (PolSAR) images with limited labeled samples, this letter proposes a memory CNN (MCNN) for semisupervised PolSAR image classification using both the labeled and unlabeled samples. Specifically, the MCNN introduces a memory module to realize an assimilation–accommodation interaction between the network and the module in the model training process. Compared with the traditional CNN-based methods, the advantage of the introduced interaction mechanism can exploit the memory information during the model training including both the learned feature representation and the model inference uncertainty. Under the framework of memory mechanism, the semisupervised learning can be implemented simply and effectively by introducing an unsupervised memory loss. We evaluate the proposed method on three benchmark PolSAR data sets. The experimental results show the advantages of the MCNN over the supervised, semisupervised, and unsupervised methods in the PolSAR image classification with limited labeled samples.

In recent years, convolutional neural networks (CNNs) have been successfully applied to inverse synthetic aperture radar (ISAR) sparse imaging because of their powerful ability in feature extraction. However, these CNNs only adopt single path feed-forward architectures and lack paths for directly transmitting original feature representations (OFRs) in shallow layers to reconstruction layers, which limits the complete reconstruction of target shape due to the underutilization of the OFRs that are efficient for recovering target details. Later, fully CNN (FCNN) introduces several skip connections (SKs) to establish the additional ways for directly passing the OFRs to the reconstruction layers. Nevertheless, the transmitted OFRs inevitably include the feature information of artifacts, which usually results the appearance of artifacts in final reconstructed target image. To address this issue, we introduce the gate units to FCNN, and refer to the improved FCNN as G-FCNN. Furthermore, the learnable gate units weight the OFRs transmitted by SKs and autonomously decide how many OFRs are transmitted further. To circumvent the shortage of the real data available for network training, we utilize the transfer learning strategy to guarantee a good performance of the G-FCNN. The imaging results of real data show that the G-FCNN-based imaging method is superior to the existing CNN-based imaging methods.

The equivalent rotation vector synthesized by 3-D nonuniform rotation of ship target leads to the time-varying characteristic of Doppler frequency of ship echo data, which will cause the image to be defocused in azimuth. In order to obtain high-resolution inverse synthetic aperture radar (ISAR) images of ship targets, we developed an optimal time selection algorithm based on time-frequency analysis of multiple scatterers. In this letter, we addressed the challenges of the serious cross-term interference in time-frequency analysis and measuring the stability of Doppler frequency. The time interval with a minimum variation of Doppler frequency was determined by using time-frequency analysis of multiple isolated scatterers and root mean squared error (RMSE). The effectiveness and robustness of the proposed algorithm were evaluated by both simulated and real ISAR data.

To overcome the problem of coarse precision in traditional estimation algorithms for estimating the three-dimensional (3D) rotational vector, this paper proposes a novel approach to estimate the total rotational vector of noncooperative targets, which uses 3D Interferometric Inverse Synthetic Aperture Radar (InISAR) technique to obtain the position estimation and effective rotational vector of the targets. Additionally, the LV's Distribution (LVD) is applied to improve the estimate precision in solving the total rotational vector along the radar line-of-sight (LOS). Finally, the total rotational vector can be obtained by combing the aforementioned two procedures. The effectiveness of the three-dimensional imaging, and the accuracy of rotation vector estimation of the proposed method are demonstrated by simulation experiments.

The traditional inverse synthetic aperture radar (ISAR) imaging uses the range-Doppler (RD) type of methods. The compressive sensing (CS)-based ISAR imaging is capable of obtaining good target images of high contrast and less sidelobe with much less downsampling data. However, the real application of CS ISAR imaging is limited by the time-consuming iteration-based image reconstruction. The image quality is also limited by the performance of sparse representation of the target scene. In recent years, deep learning methods, more specifically the convolutional neural network (CNN), has shown its capability in signal recovery with downsampling or noncomplete data. The well-trained CNN can extract high-level abstract feature representation from the input data autonomously and exploit it in the signal recovery. We are interested in exploiting the CNN to enhance the CS ISAR imaging capability. The successful training of CNN always requires many thousand annotated training samples. This limits the application of CNN to the radar imaging field where large amount of training data cannot be obtained as easy as in other fields, e.g., computer vision. We propose a fully CNN (FCNN) for ISAR imaging. The constructed FCNN has a multistage decomposition and multichannel filtering architecture and has no fully connected layers. It can work with very few training samples as compared to existing CNN-based imaging networks. The imaging results of real ISAR data show that the proposed FCNN-based ISAR imaging method outperforms the state-of-the-art CS ISAR imaging methods in both image quality and computational efficiency.

Land cover classification is an important part of the polarimetric synthetic aperture radar (PolSAR) image interpretation. The convolutional neural network (CNN) has been utilized to improve the classification accuracy recently. However, how to efficiently train the classification model with limited training samples while keeping the generalization performance is still a challenge. In this letter, we devise a subspace learning network (SSLNet) for PolSAR image classification, which can be trained more efficiently. First, a third-order polarimetric feature tensor is constructed using five-target decompositions to make full use of the prior knowledge. The tensor is then fed into a two-layer CNN in which the principal component analysis (PCA) is employed to learn the convolutional filters. Finally, the output features of the network are obtained by binary hashing and block-wise histograms, followed by the nearest neighbor (NN) classifier to complete the classification. Due to the simple learning strategy, the proposed SSLNet can be easily designed and efficiently trained. Experimental results on benchmark PolSAR data reveal that the SSLNet can achieve higher classification accuracy with limited training samples than the conventional CNN method.

Layover is an artifact observed in synthetic aperture radar (SAR) images. This artifact manifests as geometric distortions or positioning errors due to unknown or inaccurate ground topography information. Layover artifact results in erroneous distance between reconstructed scatterers, particularly in elevated regions. We develop a quantitative analysis of positioning errors due to incorrect height information in backprojection based (BP) SAR image formation. Our analysis is based on microlocal techniques and provides an explicit algebraic relationship between the two-dimensional positioning error and height error. Our analysis is applicable to arbitrary imaging geometries including bistatic configuration, arbitrary antenna trajectories, wide apertures, and large scenes. While we focus primarily on BP based image formation, the results are also applicable to range-Doppler type image formation methods. Our results can be used to interpret and identify layover regions in SAR images and extract valuable information from layover artifacts.

We consider a synthetic aperture radar (SAR) system that uses ultra-narrowband continuous waveforms (CWs) as an illumination source. Such a system has many practical advantages, such as the use of relatively simple low-cost and low-power transmitters, and in some cases, using the transmitters of opportunity, such as television and radio stations. Additionally, ultra-narrowband CW signals are suitable for motion estimation due to their ability to acquire high resolution Doppler information. In this paper, we present a novel synthetic aperture imaging method for moving targets using a bistatic SAR system transmitting ultra-narrowband CWs. Our method exploits the high Doppler resolution provided by ultra-narrowband CW signals both to image the scene reflectivity and to determine the velocity of multiple moving targets. Starting from the first principle, we develop a novel forward model based on the temporal Doppler induced by the movement of antennas and moving targets. We form the reflectivity image of the scene and estimate the motion parameters using a filtered-backprojection technique combined with a contrast optimization method. Analysis of the point spread function of our image formation method shows that reflectivity images are focused when the motion parameters are estimated correctly. We present analysis of the velocity resolution and the resolution of reconstructed reflectivity images. We analyze the error between the correct and reconstructed positions of targets due to errors in velocity estimation. Extensive numerical simulations demonstrate the performance of our method and validate the theoretical results.

We present a novel method for ground moving target imaging using a synthetic aperture radar system transmitting ultranarrowband continuous waveforms (CW). Our method exploits the high Doppler resolution provided by ultranarrowband CW signals to image both the scene reflectivity and to determine the velocity of multiple moving targets. We develop a new forward model based on the temporal Doppler induced by the movement of antennas and moving targets. The forward model relates reflectivity and velocity information at each location to a correlated received signal. We form the reflectivity images of the moving targets and estimate their motion parameters using a filtered-backprojection (FBP) technique combined with the contrast or gradient optimization method. The method results in focused reflectivity images of moving targets and their velocity estimates, regardless of the target location, speed, and velocity direction. We show that the amplitude and visible edges of the targets can be correctly reconstructed when the correct target velocity estimate is used in the FBP imaging. We present the resolution analysis of the reflectivity images. Extensive numerical simulations demonstrate the performance of our method and validate the theoretical results.

We develop a new passive imaging method for moving targets in free space using measurements from a sparse array of receivers that rely on illumination sources of opportunity. Our imaging method consists of a novel passive measurement model for moving targets and an associated image formation method. The passive measurement model for moving targets relates measurements at a given receiver to measurements at other receivers in terms of Doppler and delay based on the physics of wave propagation as well as the statistics of noise. Next, we use this model to address the image formation as a generalized likelihood ratio test for an unknown target position and velocity. The image is formed by using the position- and velocity-resolved test-statistic that is obtained by maximizing the signal-to-noise ratio of the test-statistic. When the discriminant functional is constrained to be linear, the test-statistic can be viewed as the superposition of the filtered, scaled, and delayed correlations of the measurements obtained at different receivers. We analyze the spatial and velocity resolution of the four-dimensional point spread function of our imaging method in terms of the number of receivers and transmitters and the nature of the waveforms of opportunity. We present extensive numerical simulations to demonstrate the performance of our passive moving target imaging method for different numbers of receivers and different types of waveforms of opportunity available in the real world.

We consider synthetic aperture radar (SAR) imaging using ultra-narrowband continuous waveforms (CW). Due to the high Doppler resolution of CW signals, we refer to this imaging modality as Doppler Synthetic Aperture Radar (DSAR). We present a novel model and an image formation method for the bistatic DSAR for arbitrary imaging geometries. Our bistatic DSAR model is formed by correlating the translated version of the received signal with a scaled or frequencyshifted version of the transmitted CW signal over a finite time window. High frequency analysis of the resulting model shows that the correlated signal is the projections of the scene reflectivity onto the bistatic iso-Doppler curves. We next use microlocal techniques to develop a filtered-backprojection (FBP) type image reconstruction method. The FBP inversion results in backprojection of the correlated signal onto the bistatic iso- Doppler curves as opposed to the bistatic iso-range curves, performed in the traditional wideband SAR imaging. We show that our method takes advantage of the velocity, as well as the acceleration of the antennas in certain directions to form a high resolution SAR image. Our bistatic DSAR imaging method is applicable for arbitrary flight trajectories, nonflat topography, and can accommodate system related parameters. We present resolution analysis and extensive numerical experiments to demonstrate the performance of our imaging method.

In this paper, we present a novel synthetic aperture radar imaging modality that uses ultranarrowband sources of opportunity and passive airborne receivers to form an image of the ground. Due to its combined passive synthetic aperture and high Doppler resolution of the transmitted waveforms, we refer to this modality as the Doppler Synthetic Aperture Hitchhiker or Doppler-hitchhiker for short. Our imaging method first correlates the windowed signal obtained from one receiver with the scaled and translated version of the received signal in another window from the same or another receiver. We show that this correlation processing removes the transmitter-related variables from the phase of the resulting operator that maps the radiance of the scene to the correlated signals. We define a concept of passive Doppler scale factor using the radial velocities of the receivers. Next, we show that the scaled, translated, and correlated signal is the projection of the scene radiance onto the contours that are formed by the intersection of the surfaces of constant passive Doppler scale factor and ground topography. We use microlocal analysis to design a generalized filtered-backprojection operator to reconstruct the scene radiance from its projections. Our analysis shows that the resolution of the reconstructed images improves with the increased time duration and center frequency of the transmitted ultranarrowband signals. Our reconstruction method is analytic and therefore can be made computationally efficient. Furthermore, it easily accommodates arbitrary flight trajectories, nonflat topography, and system-related parameters. We present numerical simulations to demonstrate the performance of our imaging method.

We develop a new passive image formation method capable of exploiting information about multiple scattering in the environment using measurements from a sparse array of receivers that rely on illumination sources of opportunity. We use a physics-based approach to model wave propagation and develop a statistical model that relates measurements at a given receiver to measurements at other receivers. We formulate the imaging problem as a spatially resolved binary hypothesis testing problem using the model between the measurements at different receivers, statistics of the objects to be imaged and statistics of the additive noise and clutter. We address the spatially resolved hypothesis testing problem by constraining the associated discriminant functional to be linear and by maximizing the signal-to-noise-ratio of the test-statistic, and use the resulting spatially resolved test-statistics to form the image. We present numerical simulations to demonstrate the performance of the passive imaging algorithm.

In this letter, a novel global approach to range alignment for inverse synthetic aperture radar (ISAR) image formation is presented. The algorithm is based on the minimization of the entropy of the average range profile (ARP), and the processing chain is capable of exploiting the efficiency of the fast Fourier transform. With respect to the existing global methods, the new one requires no exhaustive search operation and eliminates the necessity of the parametric model for the relative offset among the range profiles. The derivation of the algorithm indicates that the presented methodology is essentially an iterative solution to a set of simultaneous equations, and its robustness is also ensured by the iterative structure. Some alternative criteria, such as the maximum contrast of the ARP, can be introduced into the algorithm with a minor change in the entropy-based method. The convergence and robustness of the presented algorithm have been validated by experimental ISAR data.

In this work, a 2-D subaperture polar format algorithm (PFA) based on stepped-chirp signal is proposed. Instead of traditional pulse synthesis preprocessing, the presented method integrates the pulse synthesis process into the range subaperture processing. Meanwhile, due to the multi-resolution property of subaperture processing, this algorithm is able to compensate the space-variant phase error caused by the radar motion during the period of a pulse cluster. Point target simulation has validated the presented algorithm.

The images generated by a circular-scanning synthetic aperture radar (SAR) can provide precise guiding information though image-matching post-processing, which necessitates their high precision in geometry. However, due to the irregular motion of the radar platform and the circular scanning antenna beam, the inevitable geometric distortion in the focused images is necessarily to be corrected. In this paper, a two-dimensional geometric distortion correction algorithm based on projection transformation between the scatterers and the images is presented, in condition of focusing the subimages using linear range-Doppler algorithm. The geometric distortion in any subimage obtained at any squint angle within 360 degrees can be effectively corrected. The point-target simulation results are provided to demonstrate the validity of the proposed method.

Inverse Synthetic Aperture Radar (ISAR) is used to image noncooperative moving targets such as aircrafts, ships and celestial objects. The target rotation relative to the radar is the source for obtaining cross-range resolution in ISAR imaging. In airborne ISAR imaging of ships, the composition of the relative motion is more complicated than in other cases. One component is produced by the relative movement between the radar and the target. The other comes from the ship sway (roll, pitch and yaw). Furthermore, the practical sea-state changes frequently, and is also unpredictable. All of these increase the difficulty in image formation. In This work, the airborne ISAR imaging of ship targets is substantially discussed, and we design a simulation software kit applicable to practical sea-state and arbitrary flight path. By using the simulated data, the effects of the various relative rotations between the radar and the target on the ISAR image are clearly demonstrated, and this simulation work provides good experiences for further study. Finally, some imaging results under certain simulation conditions are presented.