Computational microwave imaging with metasurface aperture (MA-CMI) is an emerging real aperture imaging scheme. It is widely concerned because of its advantages of low hardware complexity, low manufacturing cost, and high sampling rate. However, the existing studies require the MA-CMI system to remain relatively stationary with the Region of Interest (ROI), which greatly limits its application. Again, if there is a relative motion between the MA-CMI system and the ROI, the correlation between the measured signal and the sensing matrix of the MA-CMI system will be lost, resulting in the defocus of the imaging results and the degradation of the imaging quality. In this paper, our work focuses on the problem that the imaging results are out of focus due to the unknown lateral motion between the MA-CMI system and the ROI, that is, the blind focus imaging of moving scenes based on the MA-CMI system. Firstly, according to the measurement mechanism of the MA-CMI system, an observation model that considers the lateral motion of the ROI is established. Subsequently, based on the model, we develop a robust alternate iterative method called Newton-Sparse Bayesian learning (Nt-SBL). Specifically, in each iteration, the Newton method was used to estimate the unknown velocity parameters of the motion scene; on the other hand, based on the current velocity estimation, the scene reconstruction was performed using the generic SBL. Finally, based on the MA-CMI system prototype we designed, both simulation and measured experiments are performed to verify the effectiveness of the proposed method.

Micro-Doppler analysis of rigid-body target is significant for attitude estimation and recognition of space objects. The traditional micro-Doppler analysis method for rigid-body targets includes two sequential steps. First, the time-frequency analysis is performed on the radar echo data of the target, and then the micro-Doppler curve of each scattering center is separated and extracted from the time-frequency map. The second step depends on the micro motion model of the target. In the micro-Doppler analysis of real rigid-body target, there are some problems such as the mismatch of the micro motion model, the incidence angle dependence of the scattering center position, and the partial occlusion of the scattering center. Therefore, it is very difficult to correctly extract the micro-Doppler curves of multiple scatterers. In this paper, an end-to-end micro-Doppler analysis method for rigid-body target based on deep learning network is proposed, which can directly separate and extract the micro-Doppler curves of multiple scatterers from the target echo data. Specifically, an Asymmetrical AutoEncoding (A2E) network equipped with a Data Pre-Processing (DP2) module is developed to extract Time-Frequency Curves (TFCs) from radar echos. Considering the sparseness of Time-Frequency Distribution (TFDs), we then develop a novel Energy-Concentration Objective (ECO) function based on Min-Max game to enhance curves energy while suppress the background energy. In practice, measured TFDs are rarely annotated, it restricts the generalization capability of the A2E from the simulation to the measurement. To bridge the gap, two-fold modifications are finally constructed: i) we insert a partial-shared branch of decoder to reconstruct the TFD from the DP2 module; ii) we regularize the ECO function with a knowledge preservation based reconstruction bound to further capture the characteristics of measured echos in a semi-supervised way at test time to relieve the domain-shift problem.

Sparse aperture (SA) autofocusing and imaging is a hot research problem in the signal processing field and has been widely used. Under SA, the absence of echoes destroys the coherence between the pulses, which then affects the autofocusing accuracy of the imaging, leading to defocus of the image. In this article, a novel SA autofocusing and imaging algorithm based on sparse Bayesian learning (SBL) is proposed, which uses a fast SBL algorithm to achieve SA high-resolution imaging and the minimum Tsallis entropy algorithm to realize autofocusing. As is known to all, SBL has strong robustness and high precision. Unfortunately, the direct calculation of the inversion and multiplication operations involved in each iteration of SBL results in significant computational costs. In the proposed fast SBL algorithm, the matrix required to be inverted has a special structure. The inverse matrix can then be represented by Gohberg–Semencul (G–S) factorization. Also, almost all operations except for G–S factorization during each iteration can be completed by fast Fourier transform (FFT) or inverse FFT (IFFT), which greatly reduces the amount of computation by several orders of magnitude. In each SBL iteration, the minimum Tsallis entropy algorithm is used for estimating the phase error, which has better noise sensitivity and obtains the images with the best focused degree. Finally, the effectiveness and high efficiency of the proposed fast algorithm are verified by experimental results obtained by simulation and measured data.

Computational microwave imaging (CMI) based on the frequency diversity metasurface apertures (FDMAs) is an emerging technology and has attracted wide attention. FDMA based CMI (FDMA-CMI) can be considered as microwave compressive sensing imaging with the frequency diversity pattern of the FDMA being the sensing matrix and solved by sparse signal reconstruction algorithms. However, the imaging quality is affected by the sensing matrix error and off-grid error seriously. In this paper, we propose a novel algorithm for FDMA-CMI, referred to as OGSISBL, by taking both the off-grid error and sensing matrix error into account. Firstly, we establish the measurement model with both the off-grid error and sensing matrix error. Specifically, the off-grid error is represented as a set of parameters to be estimated in the measurement model and the sensing matrix error is represented as the amplitude-phase drift of the transceiver channels of the imaging system due to the principle of the FDMA. Then, under the framework of the sparse Bayesian learning, a robust imaging algorithm OGSISBL is developed via the variational Bayesian expectation maximization (VBEM), which can not only recover the amplitude and position of the return of the scattered, but also simultaneously calibrate the amplitude-phase drift of the transceiver channels and the off-grid error. The performance of the proposed algorithm is evaluated by both the simulation data and the measured data collected by the self-designed experimental FDMA-CMI system, and the results validate the effectiveness and robustness of the proposed method.

Sparse Bayesian learning (SBL) is a popular and robust algorithm for sparse signal reconstruction (SSR). Unfortunately, the SBL algorithm suffers from heavy computational complexity when it is implemented directly since the inversion and multiplying operations of the estimation of the covariance matrix are involved in each iteration, which is proportional to the cube of the observed data length and thus prevents it solving large-scale problems. In many applications, such as radar imaging and array signal processing, the signal to be recovered is sparse in the Fourier dictionary. In this article, we propose an efficient implementation method for the Fourier dictionary-based SBL (FD-SBL). In the case that the Fourier dictionary is adopted, the estimation of the covariance matrix is a Toeplitz matrix for 1-D data or a Toeplitz-block-Toeplitz (TBT) matrix for 2-D data during the FD-SBL iterations. By utilizing this property, we employ the Gohberg–Semencul (G-S)-type factorization to accelerate the implementation of FD-SBL. To be noted, there is no approximation in our proposed method, and the computational cost is reduced by several orders of magnitude compared with the direct implementation of SBL. Finally, the experimental results verify the effectiveness of the proposed method.

Target detection in remote sensing image has long been one of the research focuses in related areas. This paper proposed a deep learning model for target detection in remote sensing image fusing multilevel features and applied to detect aircrafts in remote sensing images. Because the model is small in size and applies fusion of multilevel features, the detection accuracy of aircraft targets with different scales and denser in remote sensing images has been improved, without compromising the detection speed. A packet fusion reject detection bounding boxes (PFR-DBB) algorithm was also proposed, which is able to better remove redundant detection boxes and further improve detection accuracy. With the experiment results of two remote sensing aircraft data sets detection based on the model, it is proved that small-scale deep networks can also achieve high performance for multi-scale aircraft target detection on small sample data sets.

Micro-Doppler (MD) radar signatures characterize rich motion information of the targets and are of great significance in target recognition. In this letter, we propose a novel high-order particle filter track-before-detect (PF-TBD) approach for the MD curves extraction. In the proposed approach, the sinusoidal Doppler frequency curve is treated as the state, whose dynamic model is described as a high-order Markov chain. First, the state equation is divided into two parts, the translational motion part represented as a first-order dynamic process and the micromotion part represented as a high-order dynamic process including static model parameters. Then, a kernel smoothing approach is introduced for the static model parameters estimation, and the auxiliary particle filter (APF) is utilized for the instantaneous Doppler curves extraction. Finally, the experiments on the electromagnetic analysis data are carried out to validate the performance of the proposed method.

Micro-Doppler radar signatures are capable of characterizing rich motion information of targets and have played important roles in target identification and recognition. In this letter, we develop a novel parametric time-frequency method to analyze the micro-Doppler signatures of rigid-body targets, which is referred to as the block-sparse forward-backward time-varying autoregressive (BS-FBTVAR) model. First, the basis expansion method is employed to convert the time-varying model parameter estimation problem to be time invariant. Then, by investigating the intrinsic relationship between the model parameters and the poles of rigid-body targets, block-sparsity constraints are introduced to the conventional FBTVAR model. A complex-valued block-sparse Bayesian learning algorithm is developed as the solver of the novel BS-FBTVAR model. Finally, experiments on the electromagnetic (EM) analysis data are carried out to validate the performance of the proposed method.