Attention-Based Recurrent Temporal Restricted Boltzmann Machine for Radar High Resolution Range Profile Sequence Recognition

Zhang, Yifan; Gao, Xueshan; Peng, Xuan; Ye, Jiaqi; Li, Xiang

doi:10.3390/s18051585

Cited by 17 publications

(12 citation statements)

References 32 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In the step of samples extrapolation, the training sequences with missing HRRPs are chosen as input vectors. As shown in Figure 5, take a sample sequence circled in the red box on the left of the figure as an example, we decompose it into 15 independent HRRP samples (Compared with [17], we set the same sequence length. ), and then input them into successive time steps in VGM-RNN.…”

Section: B the Flow Chart Of Vgm-rnn For Sequential Hrrp Recognitionmentioning

confidence: 99%

“…In this paper, we use the widely used subset, which contains three kinds of targets with high similarity. In order to compare the recognition performance between this paper and other state-of-the-art literatures more conveniently, we convert two-dimensional SAR samples into one-dimensional HRRP sequence samples with the method consistent with literature [17]. The optical image, SAR image and single HRRP image under an azimuth of the three kinds of targets are shown in Figure 7.…”

Section: A Datasetmentioning

confidence: 99%

“…Besides, each sample in the training set contains only one model, but the BMP2 and T72 in the test set contain three models respectively, which is conducive to the detection of the generalization performance of the model, and which is very valuable for practical radar application scenarios. We use the same method as [17] to convert SAR image into HRRP sequence with length T = 15. Take the first three HRRP samples as an example, the HRRP sequence of three types of targets are shown in Figure 8, which shows that the HRRP sequences of the three kinds of targets are highly similar, but their maximum amplitudes and the range cell index of the maximum echo intensity are slightly different.…”

Section: A Datasetmentioning

confidence: 99%

“…The third and fourth group of contrast experiments adopt the deep learning method, which are Recurrent Temporal Restricted Boltzmann Machine (RTRBM) and classical RNN respectively. The number of visible layer units of the two models is set to 70 according to MSTAR data, while the number of hidden layer units is set to 128, which is consistent with the setting of [17]. In order to eliminate the random errors in the experiment, the third and fourth group of contrast experiments and the methods proposed in this paper will be repeated for five times under the same parameter settings, and the final extrapolation sample is represented by the average of five extrapolation results.…”

Section: B Experiments 1: Evaluating the Impact Of Proposed Strategiementioning

confidence: 99%

“…In Experiment 1, the number of hidden nodes in VGM-RNN is set to 128, which is consistent with [17], so as to explore the extrapolation performance of our proposed model under the same conditions.…”

Section: ) Exploring the Optimal Size Of Hidden Layer And The Value mentioning

confidence: 99%

See 4 more Smart Citations

VGM-RNN: HRRP Sequence Extrapolation and Recognition Based on a Novel Optimized RNN

et al. 2020

Self Cite

View full text Add to dashboard Cite

High-Resolution Range Profile (HRRP) sequence has attracted academic attentions in the field of radar automatic target recognition (RATR) owing to its abundant spatial-temporal correlation between adjacent samples. However, it is difficult in the working state of radar to obtain complete HRRP sequence samples due to various internal and external factors such as ground clutter and systematic error, which poses an enormous challenge to radar target recognition. Therefore, it is crucial to repair the missing HRRPs based on the adjacent samples in the previous frames. In this paper, we discuss the extrapolate method of incomplete samples and propose an improved neural network algorithm named as Vanishing Gradient Mitigation Recurrent Neural Network (VGM-RNN). The lost samples in the sequence can be extrapolated by VGM-RNN, and the problem of vanishing gradient which is possessed in classical RNN can be effectively mitigated. The proposed method in this paper can be divided into two parts, as sample extrapolation and sequence recognition, in which sample extrapolation is the core method. Experimental results on Moving and Stationary Target Acquisition and Recognition (MSTAR) dataset show that our proposed model exhibits higher accuracy and efficiency, as well as excellent anti-noise performance, compared with traditional methods. It is suggested that our proposed model can be effectively applied to radar system.

show abstract

Section: B the Flow Chart Of Vgm-rnn For Sequential Hrrp Recognitionmentioning

confidence: 99%

Section: A Datasetmentioning

confidence: 99%

Section: A Datasetmentioning

confidence: 99%

Section: B Experiments 1: Evaluating the Impact Of Proposed Strategiementioning

confidence: 99%

Section: ) Exploring the Optimal Size Of Hidden Layer And The Value mentioning

confidence: 99%

See 3 more Smart Citations

VGM-RNN: HRRP Sequence Extrapolation and Recognition Based on a Novel Optimized RNN

et al. 2020

Self Cite

View full text Add to dashboard Cite

show abstract

High‐Resolution Range Profile Sequence Recognition Based on Transformer with Temporal–Spatial Fusion and Label Smoothing

Wang,

Song

et al. 2023

Advanced Intelligent Systems

View full text Add to dashboard Cite

Radar high‐resolution range profile (HRRP) is widely used in radar automatic target recognition due to its advantages such as easy availability, convenient processing, and small storage space. Current recognition methods for HRRP sequences mainly focus on the temporal information of HRRP sequences, which cannot fully utilize the temporal and spatial information contained in HRRP sequences. Moreover, most of these methods fail in long‐range modeling and global information extraction of HRRP sequences. To solve above problems, a HRRP sequence recognition method based on transformer with temporal–spatial fusion and label smoothing (TSF–transformer–LS) is proposed. TSF–transformer–LS contains temporal transformer blocks and spatial transformer blocks, which are used to extract deep global features of HRRP sequences in the time domain and space domain, respectively. Then, an attention fusion mechanism is developed to realize the adaptive fusion of temporal and spatial features. Moreover, label smoothing is used to add noise to sample labels, which can solve the overfitting problem of transformer caused by a large amount of noise hidden in HRRP in real scenes. Experiments on MSTAR, a standard dataset, show that the proposed method outperforms other methods in recognition performance. Furthermore, the effectiveness and interpretability of the method are explored.

show abstract

Recognition of high-resolution range profile sequence based on TCN with sequence length-adaptive algorithm and elastic net regularization

Wang,

Song

et al. 2024

Expert Systems with Applications

View full text Add to dashboard Cite

Attention-Based Recurrent Temporal Restricted Boltzmann Machine for Radar High Resolution Range Profile Sequence Recognition

Cited by 17 publications

References 32 publications

VGM-RNN: HRRP Sequence Extrapolation and Recognition Based on a Novel Optimized RNN

VGM-RNN: HRRP Sequence Extrapolation and Recognition Based on a Novel Optimized RNN

High‐Resolution Range Profile Sequence Recognition Based on Transformer with Temporal–Spatial Fusion and Label Smoothing

Recognition of high-resolution range profile sequence based on TCN with sequence length-adaptive algorithm and elastic net regularization

Contact Info

Product

Resources

About