The High Resolution Range Profile (HRRP) recognition has attracted great concern in the field of Radar Automatic Target Recognition (RATR). However, traditional HRRP recognition methods failed to model high dimensional sequential data efficiently and have a poor anti-noise ability. To deal with these problems, a novel stochastic neural network model named Attention-based Recurrent Temporal Restricted Boltzmann Machine (ARTRBM) is proposed in this paper. RTRBM is utilized to extract discriminative features and the attention mechanism is adopted to select major features. RTRBM is efficient to model high dimensional HRRP sequences because it can extract the information of temporal and spatial correlation between adjacent HRRPs. The attention mechanism is used in sequential data recognition tasks including machine translation and relation classification, which makes the model pay more attention to the major features of recognition. Therefore, the combination of RTRBM and the attention mechanism makes our model effective for extracting more internal related features and choose the important parts of the extracted features. Additionally, the model performs well with the noise corrupted HRRP data. Experimental results on the Moving and Stationary Target Acquisition and Recognition (MSTAR) dataset show that our proposed model outperforms other traditional methods, which indicates that ARTRBM extracts, selects, and utilizes the correlation information between adjacent HRRPs effectively and is suitable for high dimensional data or noise corrupted data.

This paper proposes a new feature learning method for the recognition of radar high resolution range profile (HRRP) sequences. HRRPs from a period of continuous changing aspect angles are jointly modeled and discriminated by a single model named the discriminative infinite restricted Boltzmann machine (Dis-iRBM). Compared with the commonly used hidden Markov model (HMM)-based recognition method for HRRP sequences, which requires efficient preprocessing of the HRRP signal, the proposed method is an end-to-end method of which the input is the raw HRRP sequence, and the output is the label of the target. The proposed model can efficiently capture the global pattern in a sequence, while the HMM can only model local dynamics, which suffers from information loss. Last but not least, the proposed model learns the features of HRRP sequences adaptively according to the complexity of a single HRRP and the length of a HRRP sequence. Experimental results on the Moving and Stationary Target Acquisition and Recognition (MSTAR) database indicate that the proposed method is efficient and robust under various conditions.

In existing sparsity-driven inverse synthetic aperture radar (ISAR) imaging framework a sparse recovery (SR) algorithm is usually applied to azimuth compression to achieve high resolution in the cross-range direction. For range compression, however, direct application of an SR algorithm is not very effective because the scattering centers resolved in the high resolution range profiles at different view angles always exhibit irregular range cell migration (RCM), especially for complex targets, which will blur the ISAR image. To alleviate the sparse recovery-induced RCM in range compression, a sparsity-driven framework for ISAR imaging named Fourier-sparsity integrated (FSI) method is proposed in this paper, which can simultaneously achieve better focusing performance in both the range and cross-range domains. Experiments using simulated data and real data demonstrate the superiority of our proposed framework over existing sparsity-driven methods and range-Doppler methods.

The infinite restricted Boltzmann machine (iRBM) is an extension of the classic RBM. It enjoys a good property of automatically deciding the size of the hidden layer according to specific training data. With sufficient training, the iRBM can achieve a competitive performance with that of the classic RBM. However, the convergence of learning the iRBM is slow, due to the fact that the iRBM is sensitive to the ordering of its hidden units, the learned filters change slowly from the left-most hidden unit to right. To break this dependency between neighboring hidden units and speed up the convergence of training, a novel training strategy is proposed. The key idea of the proposed training strategy is randomly regrouping the hidden units before each gradient descent step. Potentially, a mixing of infinitely many iRBMs with different permutations of the hidden units can be achieved by this learning method, which has a similar effect of preventing the model from over-fitting as the dropout. The original iRBM is also modified to be capable of carrying out discriminative training. To evaluate the impact of our method on convergence speed of learning and the model’s generalization ability, several experiments have been performed on the binarized MNIST and CalTech101 Silhouettes datasets. Experimental results indicate that the proposed training strategy can greatly accelerate learning and enhance generalization ability of iRBMs.

Adversarial examples (AEs) bring increasing concern on the security of deep-learning-based synthetic aperture radar (SAR) target recognition systems. SAR AEs with perturbation constrained to the vicinity of the target have been recently in the spotlight due to the physical realization prospects. However, current adversarial detection methods generally suffer severe performance degradation against SAR AEs with region-constrained perturbation. To solve this problem, we treated SAR AEs as low-probability samples incompatible with the clean dataset. With the help of energy-based models, we captured an inherent energy gap between SAR AEs and clean samples that is robust to the changes of the perturbation region. Inspired by this discovery, we propose an energy-based adversarial detector, which requires no modification to a pretrained model. To better distinguish the clean samples and AEs, energy regularization was adopted to fine-tune the pretrained model. Experiments demonstrated that the proposed method significantly boosts the detection performance against SAR AEs with region-constrained perturbation.