Showing papers by "Shunjun Wei published in 2020"

HRSID: A High-Resolution SAR Images Dataset for Ship Detection and Instance Segmentation

Abstract: With the development of satellite technology, up to date imaging mode of synthetic aperture radar (SAR) satellite can provide higher resolution SAR imageries, which benefits ship detection and instance segmentation. Meanwhile, object detectors based on convolutional neural network (CNN) show high performance on SAR ship detection even without land-ocean segmentation; but with respective shortcomings, such as the relatively small size of SAR images for ship detection, limited SAR training samples, and inappropriate annotations, in existing SAR ship datasets, related research is hampered. To promote the development of CNN based ship detection and instance segmentation, we have constructed a High-Resolution SAR Images Dataset (HRSID). In addition to object detection, instance segmentation can also be implemented on HRSID. As for dataset construction, under the overlapped ratio of 25%, 136 panoramic SAR imageries with ranging resolution from 1m to 5m are cropped to $800 \times 800$ pixels SAR images. To reduce wrong annotation and missing annotation, optical remote sensing imageries are applied to reduce the interferes from harbor constructions. There are 5604 cropped SAR images and 16951 ships in HRSID, and we have divided HRSID into a training set (65% SAR images) and test set (35% SAR images) with the format of Microsoft Common Objects in Context (MS COCO). 8 state-of-the-art detectors are experimented on HRSID to build the baseline; MS COCO evaluation metrics are applicated for comprehensive evaluation. Experimental results reveal that ship detection and instance segmentation can be well implemented on HRSID.

LS-SSDD-v1.0: A Deep Learning Dataset Dedicated to Small Ship Detection from Large-Scale Sentinel-1 SAR Images

Abstract: Ship detection in synthetic aperture radar (SAR) images is becoming a research hotspot. In recent years, as the rise of artificial intelligence, deep learning has almost dominated SAR ship detection community for its higher accuracy, faster speed, less human intervention, etc. However, today, there is still a lack of a reliable deep learning SAR ship detection dataset that can meet the practical migration application of ship detection in large-scene space-borne SAR images. Thus, to solve this problem, this paper releases a Large-Scale SAR Ship Detection Dataset-v1.0 (LS-SSDD-v1.0) from Sentinel-1, for small ship detection under large-scale backgrounds. LS-SSDD-v1.0 contains 15 large-scale SAR images whose ground truths are correctly labeled by SAR experts by drawing support from the Automatic Identification System (AIS) and Google Earth. To facilitate network training, the large-scale images are directly cut into 9000 sub-images without bells and whistles, providing convenience for subsequent detection result presentation in large-scale SAR images. Notably, LS-SSDD-v1.0 has five advantages: (1) large-scale backgrounds, (2) small ship detection, (3) abundant pure backgrounds, (4) fully automatic detection flow, and (5) numerous and standardized research baselines. Last but not least, combined with the advantage of abundant pure backgrounds, we also propose a Pure Background Hybrid Training mechanism (PBHT-mechanism) to suppress false alarms of land in large-scale SAR images. Experimental results of ablation study can verify the effectiveness of the PBHT-mechanism. LS-SSDD-v1.0 can inspire related scholars to make extensive research into SAR ship detection methods with engineering application value, which is conducive to the progress of SAR intelligent interpretation technology.

HyperLi-Net: A hyper-light deep learning network for high-accurate and high-speed ship detection from synthetic aperture radar imagery

Abstract: Ship detection from Synthetic Aperture Radar (SAR) imagery is attracting increasing attention due to its great value in ocean. However, existing most studies are frequently improving detection accuracy at the expense of detection speed. Thus, to solve this problem, this paper proposes HyperLi-Net for high-accurate and high-speed SAR ship detection. We propose five external modules to achieve high-accuracy, i.e., Multi-Receptive-Field Module (MRF-Module), Dilated Convolution Module (DC-Module), Channel and Spatial Attention Module (CSA-Module), Feature Fusion Module (FF-Module) and Feature Pyramid Module (FP-Module). We also adopt five internal mechanisms to achieve high-speed, i.e., Region-Free Model (RF-Model), Small Kernel (S-Kernel), Narrow Channel (N-Channel), Separable Convolution (Separa-Conv) and Batch Normalization Fusion (BN-Fusion). Experimental results on the SAR Ship Detection Dataset (SSDD), Gaofen-SSDD and Sentinel-SSDD show that HyperLi-Net’s accuracy and speed are both superior to the other nine state-of-the-art methods. Moreover, the satisfactory detection results on two Sentinel-1 SAR images can reveal HyperLi-Net’s good migration capability. HyperLi-Net is build from scratch with fewer parameters, lower computation costs and lighter model that can be efficiently trained on CPUs and is helpful for future hardware transplantation, e.g. FPGAs, DSPs, etc.

Precise and Robust Ship Detection for High-Resolution SAR Imagery Based on HR-SDNet

Abstract: Ship detection in high-resolution synthetic aperture radar (SAR) imagery is a challenging problem in the case of complex environments, especially inshore and offshore scenes. Nowadays, the existing methods of SAR ship detection mainly use low-resolution representations obtained by classification networks or recover high-resolution representations from low-resolution representations in SAR images. As the representation learning is characterized by low resolution and the huge loss of resolution makes it difficult to obtain accurate prediction results in spatial accuracy; therefore, these networks are not suitable to ship detection of region-level. In this paper, a novel ship detection method based on a high-resolution ship detection network (HR-SDNet) for high-resolution SAR imagery is proposed. The HR-SDNet adopts a novel high-resolution feature pyramid network (HRFPN) to take full advantage of the feature maps of high-resolution and low-resolution convolutions for SAR image ship detection. In this scheme, the HRFPN connects high-to-low resolution subnetworks in parallel and can maintain high resolution. Next, the Soft Non-Maximum Suppression (Soft-NMS) is used to improve the performance of the NMS, thereby improving the detection performance of the dense ships. Then, we introduce the Microsoft Common Objects in Context (COCO) evaluation metrics, which provides not only the higher quality evaluation metrics average precision (AP) for more accurate bounding box regression, but also the evaluation metrics for small, medium and large targets, so as to precisely evaluate the detection performance of our method. Finally, the experimental results on the SAR ship detection dataset (SSDD) and TerraSAR-X high-resolution images reveal that (1) our approach based on the HRFPN has superior detection performance for both inshore and offshore scenes of the high-resolution SAR imagery, which achieves nearly 4.3% performance gains compared to feature pyramid network (FPN) in inshore scenes, thus proving its effectiveness; (2) compared with the existing algorithms, our approach is more accurate and robust for ship detection of high-resolution SAR imagery, especially inshore and offshore scenes; (3) with the Soft-NMS algorithm, our network performs better, which achieves nearly 1% performance gains in terms of AP; (4) the COCO evaluation metrics are effective for SAR image ship detection; (5) the displayed thresholds within a certain range have a significant impact on the robustness of ship detectors.

HQ-ISNet: High-Quality Instance Segmentation for Remote Sensing Imagery

Abstract: Instance segmentation in high-resolution (HR) remote sensing imagery is one of the most challenging tasks and is more difficult than object detection and semantic segmentation tasks. It aims to predict class labels and pixel-wise instance masks to locate instances in an image. However, there are rare methods currently suitable for instance segmentation in the HR remote sensing images. Meanwhile, it is more difficult to implement instance segmentation due to the complex background of remote sensing images. In this article, a novel instance segmentation approach of HR remote sensing imagery based on Cascade Mask R-CNN is proposed, which is called a high-quality instance segmentation network (HQ-ISNet). In this scheme, the HQ-ISNet exploits a HR feature pyramid network (HRFPN) to fully utilize multi-level feature maps and maintain HR feature maps for remote sensing images’ instance segmentation. Next, to refine mask information flow between mask branches, the instance segmentation network version 2 (ISNetV2) is proposed to promote further improvements in mask prediction accuracy. Then, we construct a new, more challenging dataset based on the synthetic aperture radar (SAR) ship detection dataset (SSDD) and the Northwestern Polytechnical University very-high-resolution 10-class geospatial object detection dataset (NWPU VHR-10) for remote sensing images instance segmentation which can be used as a benchmark for evaluating instance segmentation algorithms in the high-resolution remote sensing images. Finally, extensive experimental analyses and comparisons on the SSDD and the NWPU VHR-10 dataset show that (1) the HRFPN makes the predicted instance masks more accurate, which can effectively enhance the instance segmentation performance of the high-resolution remote sensing imagery; (2) the ISNetV2 is effective and promotes further improvements in mask prediction accuracy; (3) our proposed framework HQ-ISNet is effective and more accurate for instance segmentation in the remote sensing imagery than the existing algorithms.

JRNet: Jamming Recognition Networks for Radar Compound Suppression Jamming Signals

Abstract: As electromagnetic environments in battlefields are more and more complex, there are more kinds of suppression jamming noise including both single jamming signals and compound jamming signals. Thus, suppression jamming recognition especially for compound jamming signals is becoming vital and challenging. Whereas conventional methods are prone to owning the low recognition accuracy and the high computational complexity, especially under low jamming-to-noise ratio (JNR) conditions. In this paper, a novel jamming recognition network (JRNet) based on robust power-spectrum features is proposed to recognize ten kinds of suppression jamming signals including four single jamming patterns and six compound jamming patterns. The proposed method combines the significant representative power of the JRNet and distinguished power-spectrum features of jamming signals to promote the recognition performance. By integrating residual blocks and asymmetric convolution blocks, the JRNet is capable to address the degradation problem and enhance the recognition ability for subtle features. The simulation results show that the overall recognition accuracy of the proposed method is more than 90% even when the JNR is −18 dB and is close to 100% at −6 dB. Compared with five comparison methods in recent literatures, the proposed JRNet achieves better and stable recognition performance especially under low JNR conditions with relatively less storage source and a bit more FLOPs and inference time.

Intra-pulse modulation radar signal recognition based on CLDN network

Abstract: Automatic modulation classification of radar signals, which plays a significant role in both civilian and military applications, is researched in this study through a deep learning network. In this study, a novel network combined a shallow convolution neural network (CNN), long short-term memory (LSTM) network and deep neural network (DNN) is proposed to recognise six types of radar signals with different signal-to-noise ratio (SNR) levels from -14 to 20 dB. First, raw signal sequences in the time domain, frequency domain and autocorrelation domain are as input for a shallow CNN. Then the features extracted by CNN will be the input of LSTM network. Finally, DNNs will output the signal modulation types directly. The simulation results demonstrate that the accuracies in autocorrelation domain are all more than 90% at -6 dB and close to 100% when SNR > -2 dB. The recognition performances of the three domains are compared. Compared with other recognition methods, the proposed method has higher average accuracy and better performance under low SNR condition. The measured results show that the proposed method has achieved high accuracies of common four kinds of measured radar signals.

CSR-Net: A Novel Complex-Valued Network for Fast and Precise 3-D Microwave Sparse Reconstruction

Abstract: Since the compressed sensing (CS) theory broke through the limitation of the traditional Nyquist sampling theory, it has attracted extensive attention in the field of microwave imaging. However, in 3-D microwave sparse reconstruction application, conventional CS-based algorithms always suffer from huge computational cost. In this article, a novel 3-D microwave sparse reconstruction method based on a complex-valued sparse reconstruction network (CSR-Net) is proposed, which converts complex number operations into matrix operations for real and imaginary parts. Using the unfolding + network approximate scheme, each iteration process of CS-based iterative threshold optimization is designed as a block of CSR-Net, and a modified shrinkage term is introduced to improve the convergence performance of the approach. In addition, CSR-Net adopts a convolutional neural network module to replace a nonlinear sparse representation process, which dramatically reduces computational complexity and improves reconstruction performance over conventional CS-based iterative threshold optimization algorithms. Then, we divide the 3-D scene into a series of 2-D slices, and a phase correction scheme is adopted to ensure that the whole 3-D scene can be reconstructed with measurement matrix of a slice. Moreover, an efficient position–amplitude–random training method without additional real-measured data is employed for the proposed network, which effectively train the CSR-Net without enough real-measured data. Extensive experiment results demonstrate that CSR-Net outperforms both conventional iterative threshold optimization methods and deep network-based ISTA-NET-plus large margins. Its speed and reconstruction accuracy in 3-D imaging can achieve a state-of-the-art level.

Ground Moving Target Tracking and Refocusing Using Shadow in Video-SAR

Abstract: Stable and efficient ground moving target tracking and refocusing is a hard task in synthetic aperture radar (SAR) data processing. Since shadows in video-SAR indicate the actual positions of moving targets at different moments without any displacement, shadow-based methods provide a new approach for ground moving target processing. This paper constructs a novel framework to refocus ground moving targets by using shadows in video-SAR. To this end, an automatic-registered SAR video is first obtained using the video-SAR back-projection (v-BP) algorithm. The shadows of multiple moving targets are then tracked using a learning-based tracker, and the moving targets are ultimately refocused via a proposed moving target back-projection (m-BP) algorithm. With this framework, we can perform detecting, tracking, imaging for multiple moving targets integratedly, which significantly improves the ability of moving-target surveillance for SAR systems. Furthermore, a detailed explanation of the shadow of a moving target is presented herein. We find that the shadow of ground moving targets is affected by a target’s size, radar pitch angle, carrier frequency, synthetic aperture time, etc. With an elaborate system design, we can obtain a clear shadow of moving targets even in X or C band. By numerical experiments, we find that a deep network, such as SiamFc, can easily track shadows and precisely estimate the trajectories that meet the accuracy requirement of the trajectories for m-BP.

PRI Modulation Recognition Based on Squeeze-and-Excitation Networks

Abstract: Pulse repetition interval (PRI) modulation recognition is a significant means to analyze radar working statuses and missions in Electronic Support system. Traditional methods may be insufficient to accurately recognize complex PRI at low SNR with high percentages of missing pulses. In this letter, an approach based on Squeeze-and-Excitation networks (SE-Net) and the autocorrelation functions for PRI modulation recognition automatically is proposed. Firstly, the features of six PRI modulation types in the autocorrelation domain are converted into images by calculating instantaneous autocorrelation functions. Then, the images will be the input of SE-Net which will automatically learn about deep features of different PRI modulation modes. Finally, SE-Net will output PRI modulation modes directly. Simulation results show that SE-Net is robust to the noise and missing pulses. The accuracies for all PRI modulation modes are more than 95% at −10dB with 30% missing pulses and more than 96% at −2dB with 50% missing pulses. Compared with traditional method and other networks, SE-Net can achieve better recognition performance at low SNR and high percentages of missing pulses.

A Phase Filtering Method with Scale Recurrent Networks for InSAR

Abstract: Phase filtering is a key issue in interferometric synthetic aperture radar (InSAR) applications, such as deformation monitoring and topographic mapping. The accuracy of the deformation and terrain height is highly dependent on the quality of phase filtering. Researchers are committed to continuously improving the accuracy and efficiency of phase filtering. Inspired by the successful application of neural networks in SAR image denoising, in this paper we propose a phase filtering method that is based on deep learning to efficiently filter out the noise in the interferometric phase. In this method, the real and imaginary parts of the interferometric phase are filtered while using a scale recurrent network, which includes three single scale subnetworks based on the encoder-decoder architecture. The network can utilize the global structural phase information contained in the different-scaled feature maps, because RNN units are used to connect the three different-scaled subnetworks and transmit current state information among different subnetworks. The encoder part is used for extracting the phase features, and the decoder part restores detailed information from the encoded feature maps and makes the size of the output image the same as that of the input image. Experiments on simulated and real InSAR data prove that the proposed method is superior to three widely-used phase filtering methods by qualitative and quantitative comparisons. In addition, on the same simulated data set, the overall performance of the proposed method is better than another deep learning-based method (DeepInSAR). The runtime of the proposed method is only about 0.043s for an image with a size of 1024×1024 pixels, which has the significant advantage of computational efficiency in practical applications that require real-time processing.

CIST: An Improved ISAR Imaging Method Using Convolution Neural Network

Abstract: Compressive sensing (CS) has been widely utilized in inverse synthetic aperture radar (ISAR) imaging, since ISAR measured data are generally non-completed in cross-range direction, and CS-based imaging methods can obtain high-quality imaging results using under-sampled data. However, the traditional CS-based methods need to pre-define parameters and sparse transforms, which are tough to be hand-crafted. Besides, these methods usually require heavy computational cost with large matrices operation. In this paper, inspired by the adaptive parameter learning and rapidly reconstruction of convolution neural network (CNN), a novel imaging method, called convolution iterative shrinkage-thresholding (CIST) network, is proposed for ISAR efficient sparse imaging. CIST is capable of learning optimal parameters and sparse transforms throughout the CNN training process, instead of being manually defined. Specifically, CIST replaces the linear sparse transform with non-linear convolution operations. This new transform and essential parameters are learnable end-to-end across the iterations, which increases the flexibility and robustness of CIST. When compared with the traditional state-of-the-art CS imaging methods, both simulation and experimental results demonstrate that the proposed CIST-based ISAR imaging method can obtain imaging results of high quality, while maintaining high computational efficiency. CIST-based ISAR imaging is tens of times faster than other methods.

ACSE Networks and Autocorrelation Features for PRI Modulation Recognition

Abstract: Pulse repetition interval (PRI) modulation recognition plays an important role in electronic warfare. Conventional recognition methods based on handcrafted features and elaborate threshold values suffer from the accuracy for multiple PRI modulations at low signal-to-noise ratio (SNR) with high percentages of missing pulses. In this letter, a method based on Asymmetric Convolution Squeeze-and-Excitation (ACSE) networks and features in autocorrelation domain is proposed to recognize six PRI modulation modes automatically. First, features in the time domain, frequency domain and autocorrelation domain are converted to images. Then the images are input into ACSE networks which can extract and learn deep features without complex data pre-processing. Finally, a linear layer will output modulation modes directly. Via simulations, robustness of autocorrelation features is proved. The simulation results also demonstrate that the proposed recognition method can achieve higher than 91% accuracy at −12 dB under normal conditions for six modulations and higher than 95% at −4 dB under extreme conditions. Compared with the conventional SVM and three CNN methods, ACSE networks outperform at low SNRs under extreme conditions.

Automatic Modulation Recognition for Radar Signals via Multi-Branch ACSE Networks

Abstract: Automatic modulation recognition (AMR) for radar signals plays a significant role in electronic warfare. Conventional recognition methods may suffer from the recognition accuracy and the computation complexity under low signal-to-noise ratio (SNR) conditions. In this paper, a novel multi-branch Asymmetric Convolution Squeeze-and-Excitation (ACSE) networks using multi-domain features and fusion strategy based on a support vector machine is proposed to recognize eight kinds of radar signals. First, features of radar signals in the frequency domain, the autocorrelation domain, and the time-frequency domain are extracted. Then the obtained multi-domain features are converted as the input of the proposed networks which owns the representational power and learning ability. Finally, the outputs of multi-branch ACSE networks are fused via the fusion strategy to obtain the final results. Via simulations, the robustness and effectiveness of the fusion strategy are verified. The results on the simulation dataset prove that the proposed method can achieve more than 93% accuracy at -10dB for all modulations. Compared with four newly proposed networks, the multi-branch ACSE networks achieves better performance under low SNR conditions. And the results on measured signals show that the proposed method outperforms other comparison methods, especially for binary frequency-shift keying (BFSK) signals.

Intra-pulse modulation radar signal recognition based on Squeeze-and-Excitation networks

Abstract: In this paper, a novel recognition method based on the squeeze-and-excitation networks (SE-Nets) is proposed in order to recognize intra-pulse modulation signals at varying noise levels automatically. Firstly, different signal transforms including time domain, frequency domain and time–frequency domain are used to convert seven different intra-pulse modulation signals into images. Then, since the SE-Net has great advantages in image processing, the images are classified by the squeeze-and-excitation networks and output the results in each domain. Lastly, the decisions of different domains are combined to obtain the final results. The simulation results demonstrate that the recognition accuracies are all more than 95% except BPSK signals which are still more than 90% at the case of − 8 dB. Compared with several other neural networks and traditional support vector machine method, the accuracy of SE-Net method has increased more than 2% and performs better under different SNR conditions. The measured signals results show that the accuracies of the SE-Net method are higher than those of several other neural networks, especially for BASK and BFSK signals.

Balanced Feature Pyramid Network for Ship Detection in Synthetic Aperture Radar Images

Abstract: Ship detection in Synthetic Aperture Radar (SAR) images is a fundamental but challenging task. Nowadays, given that the huge imbalance between sparse-distribution ships and complex backgrounds in training process, most existing deep-learning-based SAR ship detectors often face great difficulty in further improving accuracy. Therefore, to solve this problem, in this paper, a novel Balanced Feature Pyramid Network (B-FPN) is applied to enhance detection accuracy. Different from the raw Feature Pyramid Network (FPN), B-FPN utilizes the same-deep integration balanced semantic features to strengthen the multi-level features in the feature pyramid, by means of four steps, namely rescaling, integrating, refining and strengthening, which do not increase too much network parameter quantity. Experimental results on the open SAR Ship Detection Dataset (SSDD) shows that B-FPN can make a 7.15% mean Average Precision (mAP) improvement than FPN.

Unambiguous Reconstruction for Multichannel Nonuniform Sampling SAR Signal Based on Image Fusion

Abstract: Multichannel signal processing in azimuth is a vital technique to enable a wide-swath Synthetic Aperture Radar (SAR) with high azimuth resolution. However, the multichannel high-resolution and wide-swath (HRWS) SAR system always suffers from the problem of the azimuth nonuniform sampling resulting in the image ambiguity, when it does not satisfy the uniform sampling condition. In this paper, to suppress the azimuth image ambiguity, we propose a novel unambiguous reconstruction method based on image fusion. During this reconstruction processing, the Back Projection (BP) algorithm is first utilized for SAR imaging to obtain the designed sub-images. Then, the reconstruction expression is derived as the summation of the sub-images weighted by the interpolation coefficient. This method integrates the reconstruction into the imaging process and the image fusion makes the procedure simple. In addition, the interpolation period, which affects the reconstruction image quality and efficiency, is further analyzed. Moreover, as the curved trajectory platform brings more challenges for the unambiguous reconstruction, the performance of the proposed method applied to the curved trajectory platform is studied. Finally, experimental results clearly verify the effectiveness of the proposed method for ambiguity suppression and demonstrate its applicability to the curved trajectory.

Shipdenet-18: An Only 1 Mb With Only 18 Convolution Layers Light-Weight Deep Learning Network For Sar Ship Detection

Abstract: With the rise of Artificial Intelligence (AI), many previous studies have already applied Deep Learning (DL) for ship detection from Synthetic Aperture Radar (SAR) imagery. However, these network scale and model size are both rather huge, leading to more computation costs. As a result, ship detection speed is bound to decline due to more computation costs, and FPGA/DSP transplantation also becomes more challenging coming from huge mode size. Therefore, to solve these problems, this paper proposes a novel lightweight deep learning network for SAR ship detection named ShipDeNet-18 (only 18 convolution layers). Essentially, fewer layers and fewer kernels jointly contribute to ShipDeNet-18's light-weight characteristic. In addition, to compensate for the severe detection accuracy's sacrifice, we also propose a Deep and Shallow Feature Fusion Module (DSFF-Module) and a Feature Pyramid Module (FP-Module), which can effectively improve its detection accuracy. Experimental results on the open SAR Ship Detection Dataset (SSDD) reveal that ShipDeNet-18's detection speed is largely superior to the other state-of-the-art detectors, meanwhile its detection accuracy is only slightly inferior to others. ShipDeNet-18 is a brand-new deep learning network built from scratch, more light-weight than the other detectors, with fewer parameters (228,246), lower computation costs (456,042 FLOPs), and smaller model size (1 MB). It is of great value in some real-time SAR application, and is also convenient for future hardware transplantation (FPGA/DSP).

A Novel Ground Moving Target Radial Velocity Estimation Method for Dual-Beam Along-Track Interferometric Sar

Abstract: Traditional dual-beam along-track interferometric synthetic aperture radar(DB-AT-InSAR) system uses along-track inter-ferometry(ATI) technique to calculate the radial velocity. However, the interferometric phase acquired by ATI can be easily affected by the noise and static clutter, which may reduce the accuracy of radial velocity estimation. In this paper, a novel ground moving target radial velocity estimation method is proposed for DB-AT-InSAR. First, the azimuth squint angle of the DB-AT-InSAR is reduced to make the fore and aft beams overlap. Then, only the overlapping subaperture echoes are used to reconstruct the SAR images by back projection algorithm. Finally, clutter suppression interferome-try(CSI) technique is applied to acquire high-accuracy radial velocity estimation. Since CSI technique can be able to suppress the clutter, it can obtain better interferometric phase than ATI, which improves the accuracy of radial velocity estimation. Simulation results show that the proposed method can obtain higher accuracy in radial velocity estimation than ATI method for DB-AT-InSAR system.

Semi-Supervised Learning-Based Remote Sensing Image Scene Classification Via Adaptive Perturbation Training

Abstract: Deep neural networks have been widely applied and researched in remote sensing image scene classification and achieved a great success. However, deep supervised network heavily relies on a large amount of labeled data. The annotation is difficult and time-consuming to obtain but the unlabeled data are comparably easier to get. Considering that, we introduce a semi-supervised learning framework for remote sensing image scene classification. The network is trained by a novel adaptive perturbation training method. The experiments on NWPU-RESISC45 dataset prove that the introduced semi-supervised classification method can achieve higher classification accuracy with unlabeled data compared with the corresponding supervised classifier, and the designed adaptive perturbation training can further improve the performance of the semi-supervised learning-based classification network.

FDBP-InSAR: An Efficient Algorithm for InSAR Imaging via Frequency Domain Back Projection

Abstract: High-quality focusing with accurate phase-preserving is a significant and challenging step in interferometric synthetic aperture radar (InSAR) imaging. Compared with conventional frequency-based imaging algorithms, the time-domain back-projection algorithm (TDBPA) can greatly ensure the accuracy of imaging and phase-preserving by point-to-point coherent integration but suffers from huge computational complexity. In this paper, we propose an efficient InSAR imaging method, called a frequency-domain back-projection algorithm (FDBPA), to achieve high-resolution focusing and accurate phase-preserving of InSAR imaging. More specifically, FDBPA is utilized to replace the traditional point-to-point coherent integration of TDBPA with frequency-domain transform. It divides the echo spectrum into uniform grids and transforms the range compression data into the range frequency domain. Phase compensation and non-uniform Fourier transform of the underlying scene are implemented to achieve image focusing in the wavenumber domain. Then, the interferometric phase of the target scene can be preserved by accurate phase compensation of the target’s distance. FDBPA avoids the repetitive calculation of index values and point-to-point coherent integration which reduces the time complexity compared with TDBPA. The characteristics of focusing and phase-preserving of our method are analyzed via simulations and experiments. The results demonstrate the efficiency and high-quality imaging of the FDBPA method. It can improve the imaging efficiency by more than three times, while keeping similar imaging accuracy compared with TDBPA.

ISAR Compressive Sensing Imaging Using Convolution Neural Network with Interpretable Optimization

Abstract: Compressive Sensing(CS) has been widely utilized in Inverse synthetic aperture radar(ISAR) imaging since real ISAR data is easier to be non-completed, and CS-based methods can obtain high-quality imaging results using under-sampled data. However, traditional CS-based methods need pre-defined parameters, sparse transforms and iterative reconstruction processes. Optimal parameters as well as transforms are tough to be hand-crafted, and iterative reconstruction consumes plenty of time, which limit practical applications in ISAR imaging. Given that Convolution Neural Network(CNN) has great power to learn rapidly, we compose CNN with traditional Iterative Shrinkage-Thresholding Algorithm(ISTA) to propose CNN-ISTA(CIST)-based ISAR imaging method. CIST is capable of learning optimal parameters and transforms throughout the training (i.e. the optimization process is interpretable) instead of manually defined. Compared with traditional state-of-the-art CS imaging methods, the experimental results demonstrate that our proposed CIST-based imaging method is superior in both imaging quality and computational efficiency.

Linear Array 3-D SAR Sparse Imaging via Convolutional Neural Network

Abstract: Compressed sensing theory has attracted extensive attention in the field of linear array 3-D Synthetic Aperture Radar (SAR) sparse imaging. However, conventional CS-based algorithms always suffer from quite huge computational cost. In this paper, we propose a new method for 3-D SAR sparse imaging based on convolutional neural network (CNN). Inspired by the work of ISTA-NET, a complex-valued version for imaging tasks is modified. Furthermore, we introduce a approximate phase correction scheme for 3-D imaging, it makes the proposed method works with only a constant measurement matrix corresponding to any slice. Moreover, Using a random training strategy, ISTA-NET networks for 3-D SAR imaging are effectively trained. Experimental results demonstrate that the proposed method outperforms conventional ISTA large margins in both accuracy and speed.

Kernel Rotational Network for Synthetic Aperture Radar Target Recognition

Abstract: Convolutional Neural Networks (CNNs) have excellent ability in image recognition, however, the requirement of a large amount of labeled dataset limits its application in the field of synthetic aperture radar (SAR) image processing. In this paper, a kernel rotational network (KR-Net) for SAR target recognition is constructed. When the labeled dataset is small, the KR-net can achieve higher classification rate than standard CNNs benefit from its inherent rotational convolution units. Also, weights sharing strategy is introduced to increase network capacity without multiplying the number of weights parameters. Meanwhile, a simple and feasible multi-branch feature converging method for the KR-Net is proposed to fuse features of rotational convolution units. Experimental results show that our network can achieve state-of-art result in the MSTAR dataset, especially when the training set is small.

Efficient Insar Imaging Based on Frequency-Domain Back Projection Algorithm

Abstract: High resolution imaging of interferometric synthetic aperture radar (InSAR) usually requires fine focusing and phase-preserving. Time-domain back projection (TDBP) method outperforms other conventional methods at focusing and phase-preserving, but suffer from huge computational complexity when the underlying scene is large. In this article, an efficient method exploiting by frequency-domain back projection (FDBP) is presented for high-resolution InSAR imaging. In the scheme, the coherent integration of focusing is efficient achieved by frequency-domain Fourier transform, and a delayed-distance is compensated to phase-preserving of InSAR. Simulation and experiment results demonstrates that FDBP algorithm improves the computational efficiency by three times while maintaining the similar focusing accuracy compared with the conventional TDBP method.

Balance Scene Learning Mechanism for Offshore and Inshore Ship Detection in SAR Images

Abstract: Huge imbalance of different scenes' sample numbers seriously reduces Synthetic Aperture Radar (SAR) ship detection accuracy. Thus, to solve this problem, this letter proposes a Balance Scene Learning Mechanism (BSLM) for offshore and inshore ship detection in SAR images.