There are several major challenges in detecting and recognizing multiple hidden objects from millimeter wave SAR security inspection images: inconsistent clarity of objects, similar objects, and complex background interference. To address these problems, a suspicious multi-object detection and recognition method based on the Multi-Path Extraction Network (MPEN) is proposed. In MPEN, You Only Look Once (YOLO) v3 is used as the base network, and then the Multi-Path Feature Pyramid (MPFP) module and modified residual block distribution are proposed. MPFP is designed to output the deep network feature layers separately. Then, to distinguish similar objects more easily, the residual block distribution is modified to improve the ability of the shallow network to capture details. To verify the effectiveness of the proposed method, the millimeter wave SAR images from the laboratory’s self-developed security inspection system are utilized in conducting research on multi-object detection and recognition. The detection rate (probability of detecting a target) and average false alarm (probability of error detection) rate of our method on the target are 94.6% and 14.6%, respectively. The mean Average Precision (mAP) of recognizing multi-object is 82.39%. Compared with YOLOv3, our method shows a better performance in detecting and recognizing similar targets.

For real-time millimeter-wave (MMW) imaging applications, spatially 2-D sparse sampling is one of the most promising techniques to reduce the heavy cost of signal acquisition. Therefore, in this article, we develop a new MMW synthetic aperture radar (SAR) sparse imaging method by establishing a 2-D pseudorandom spiral-sampling pattern. The method consists of two phases: sparse sampling-pattern design and sparse-imaging processing. The sparse sampling-pattern design is modeled as an optimization programming problem, in which the initial sampling pattern is a classical Archimedes spiral, and then, the number of samples per cycle is adjusted by coming up with a modified sigmoid function. This function is employed to increase the angular aperiodicity and randomness over the whole planar aperture. For sparse-imaging processing, a singular value thresholding (SVT) algorithm is applied to reconstruct the complete echo from partial samples observed by sparse sampling, and then, the inversion can be performed via the traditional imaging algorithms. In particular, a performance-guarantee condition of SVT is introduced to control the minimum number of samples, which bridges the relationship between pattern optimization and reconstruction performance. Our motivation is to answer what the undersampling pattern is and to estimate the extent to which the number of samples can be reduced by the current sparse imaging method. Finally, simulations and experiments validate the feasibility and effectiveness of the proposed sampling pattern, and the imaging results further demonstrate that our method with partially observed samples can offer the performance comparable to dense sampling.

High-frequency vibration of motion platform leads to paired echo for synthetic aperture radar (SAR) imaging, especially in terahertz band due to its shorter wave length. Different from most existing parameters estimation methods only considering single component high-frequency vibration, in this paper a novel method considering multi-components vibration model is proposed based on fractional Fourier transform (FrFT) with combination of quasi-maximum likelihood (QML) and random sample consensus (RANSAC). Based on the model establishment of high-frequency vibration error in the echo, its instantaneous chirp rate (ICR) is firstly estimated by FrFT in sliding sub-aperture, followed which the vibration parameters are coarsely obtained through spectrum analysis and least square (LS) regression. To further refine the parameters estimates, QML is developed for compensating the deviation both caused by the frequency spectrum leakage and the error propagation effects by one-dimensional search over the vibration frequency. Meanwhile, RANSAC is adopted for avoiding the outlier of the ICR estimates in LS regression, especially at low signal-to-noise ratio (SNR). Thus, the refinement strategy based on the combination of QML and RANSAC is developed, whose utilization improves the estimation accuracy of vibration parameters. Finally, the paired echo in terahertz SAR (THz-SAR) imaging is effectively suppressed by the proposed method, and the high-quality THz-SAR imaging results are achieved. Both simulations of single component and multi-components high-frequency vibration are used to verify the validity of the proposed method. The simulation results show that the proposed method has higher estimation accuracy even at low SNR.

Compared with microwave synthetic aperture radar (SAR), terahertz SAR (THz-SAR) is easier to achieve ultrahigh-resolution image due to its higher frequency and shorter wavelength. However, higher carrier frequency makes THz-SAR image quality very sensitive to high-frequency vibration error of motion platform. Therefore, this paper proposes a novel high-frequency vibration error estimation and compensation algorithm for THz-SAR imaging based on local fractional Fourier transform (LFrFT). Firstly, the high-frequency vibration error of the motion platform is modeled as a simple harmonic motion and THz-SAR echo signal received in each range pixel can be considered as a sinusoidal frequency modulation (SFM) signal. A novel algorithm for the parameter estimation of the SFM signal based on LFrFT is proposed. The instantaneous chirp rate of the SFM signal is estimated by determining the matched order of LFrFT in a sliding small-time window and the vibration acceleration is obtained. Hence, the vibration frequency can be estimated by the spectrum analysis of estimated vibration acceleration. With the estimated vibration acceleration and vibration frequency, the SFM signal is reconstructed. Then, the corresponding THz-SAR imaging algorithm is proposed to estimate and compensate the phase error caused by the high-frequency vibration error of the motion platform and realize high-frequency vibration error estimation and compensation for THz-SAR imaging. Finally, the effectiveness of the novel algorithm proposed in this paper is demonstrated by simulation results.

Automatic ship target detection technique is a typical and meaningful application for synthetic aperture radar (SAR) image interpretation. Nevertheless, the detection of ship targets within SAR imagery is encumbered by several detracting elements, including obscured outlines, varying dimensions, and elaborate backgrounds, which collectively render the identification process challenging. Existing methodologies for discerning ship targets prove inadequate in effectively navigating these complications. Therefore, we propose a new deep neural network to automatically detection ship target from SAR images, which is named as Hierarchical Feature Fusion and Attention Network (HFFANet). HFFANet is based on CSPDarknet, the backbone network of YOLOX, and Adaptive Feature Fusion and Attention (AFFA) module is innovated to enhance feature extraction. In AFFA, Adaptive Multi-level Feature Fusion (AMFF) module is proposed to achieve effective multi-level feature adaptive fusion to better extract target contours and suppress background clutter to reduce false alarms, and Enhanced Residual Coordinate Attention (ERCA) module is also proposed to enhance spatial location information and embed it into channel features in the channel layer. The experiments on SAR ship data set have been conducted, and the mean Average Precision (mAP) of HFFANet is 98.53%. Compared with the classical networks, the experimental results show that our model not only achieves the optimal balance in Precision and Recall, but also achieves the optimal calculation cost.

In practical applications, airborne terahertz synthetic aperture radar (THz-SAR) echo signal is not only affected by range amplitude and phase errors introduced by non-ideal THz device, but also affected by motion error in azimuth induced by the non-ideal trajectory of platform, resulting in the two-dimensional (2-D) defocus of image. To address the above errors, a novel 2-D autofocusing algorithm for real airborne stripmap THz-SAR imaging is proposed. In range autofocusing, the amplitude and phase errors are estimated simultaneously in range frequency domain based on the dominant point target selected in the preliminary focused images. To avoid dividing more sub-apertures in azimuth error estimate, the autofocusing method based on the maximization of image contrast is adopted. Then, range-Doppler algorithm integrated with the proposed 2-D autofocusing are performed to obtain the focused images. The real measured data, acquired by a 0.22 THz airborne SAR system, are used to demonstrate the validity of the proposed algorithm. By comparing the imaging results, the proposed method can effectively and efficiently focus airborne stripmap THz-SAR images.