Most existing adversarial attack methods for remote sensing images merely add adversarial perturbations or patches, resulting in visually unnatural modifications. Clouds are common atmospheric effects in remote sensing images. Generating clouds on these images can produce adversarial examples better aligning with human perception. In this paper, we propose an adversarial attack framework that leverages natural cloud patterns as perturbations. Common Perlin noise-based cloud generation is a random, non-optimizable process, which cannot be directly used to attack the target models. We design a Perlin Gradient Generator Network (PGGN), which takes a compact gradient parameter vector (gradient vectors, coefficients, and scaling factors) as input and generates multiscale Perlin noise gradient grids. Through hierarchical computations, these grids produce scale-specific cloud masks, which are adaptively fused via learnable mixing coefficients and scaling factors. Crucially, the entire cloud generation process is formulated as a black-box optimization problem, where the cloud parameter vector is iteratively refined using the Differential Evolution (DE) algorithm. This approach enables query-efficient black-box attacks by directly aligning cloud shapes with adversarial objectives, while preserving natural cloud textures. Comprehensive experiments demonstrate the strong attack capabilities of this method, along with its high query efficiency. Furthermore, we conduct an in-depth analysis of the transferability of the generated adversarial examples and their robustness in adversarial defense scenarios.

Classification of hypersaline water bodies, e.g., salt fields and salt lakes, presents unique challenges due to the similar spectral and scattering characteristics of various saline water bodies. To address this issue, we propose an innovative classification method tailored for such environments, integrating Sentinel-1 multitemporal polarimetric synthetic aperture radar (PolSAR) data with Landsat multispectral imagery. The method introduces a novel PolSAR-based feature, termed scattering mechanism entropy, to quantify variations in salt crystal precipitation processes. Additionally, the blue band from multispectral imagery is leveraged to represent ion concentrations in hypersaline water bodies. By deriving the statistical relationship between scattering mechanism entropy and blue band data for each pixel, we amplify the separability of salt features and mitigate the influence of spectral similarity. These derived features are then concatenated into a Chernoff distance-based classifier for improved classification performance. To validate the robustness and effectiveness of this method, we apply it to the classification of salt fields and salt lakes on four major salt lake sites in China: Qarhan salt lake (2018–2020), Yiliping salt lake (2021–2023), Taijnar salt lake (2019–2023), and Gasikule salt lake (2019–2023). The proposed approach achieves classification accuracies of 90.91%, 92.73%, 97.40%, and 97.65%, respectively, significantly outperforming existing water body classification methods.

Deploying deep learning-based remote sensing image interpretation models in orbit can alleviate data downlink pressure and enhance processing efficiency. To adapt to the arithmetic conditions of on-orbit platforms, deep learning models must undergo compilation before deployment, where the search and measurement of operator scheduling are critical aspects. The existing compilation methods related to the Tensor Virtual Machine (TVM) compiler suffer from inefficiencies caused by suboptimal starting point selection and extensive search spaces during the search process. Additionally, the clustering algorithm employed during the measurement of candidate scheduling relies on fixed dimensions and lacks flexible application management, resulting in slow convergence. This paper proposes corresponding optimization methods built upon the TVM framework to address these two challenges. In the search aspect, techniques such as starting point planning, salient reduction, and model sharing considerably reduce search time. In the measurement aspect, the clustering process is optimized through adaptive scheduling management and dimension adaptation, enhancing measurement efficiency. Experimental results indicate that, compared to current state-of-the-art deep learning compilers, the proposed method increases compilation speed by an average of 2.6 times while maintaining consistent inference speed.

Phase filtering based on deep learning is currently one of the hot research directions in the processing of interferometric SAR data. In this paper, a filtering model is designed using a deep learning framework and the non-local characteristics of interferometric phase maps. The method divides the interferometric phase map into image patches, with each undergoing feature extraction via structurally identical, weight-sharing encoders and decoders to leverage the map’s non-local similarities. Experimental outcomes show this approach outperforms conventional spatial-domain, transform-domain, and deep learning-based filtering methods.

Compact Polarimetric (CP) SAR has a larger imaging bandwidth compared to full-polarimetric SAR and more polarization information than dual-polarimetric SAR, which has significant potential in maritime ship detection. The non-uniform scale of ship targets and complex backgrounds present detection challenges. This paper proposes an improvement to YOLOv8 by employing deformable convolutions in the backbone network and adding attention mechanisms in the network neck. Deformable convolutions excel at extracting multi-scale features of ships with strong expressive capability, while attention mechanisms suppress the learning of background features. The paper utilizes a CP SAR dataset constructed by using high-information-content SPAN images and augments the dataset. In comparison with the experimental results of the standard YOLOv8 model, our method demonstrates an improvement of 3.2% in recall, 3.6% in precision, and 1.5% in mAP. The results indicate the effectiveness of our approach in the task of ship detection using CP SAR.

SAR ship detection is an important technology supporting water traffic monitoring and marine safety maintenance. In recent years, many methods based on deep neural networks have been used to improve the performance of SAR ship detection. These methods mainly focus on two issues: one is the false alarm of ship detection in complex inshore environments, and the other is the effective extraction and utilization of SAR ship features. The topic discussed in this paper is one of the culprits that has caused the aforementioned two problems, but has long been overlooked. Specifically, it pertains to the issue of intra-class imbalance in SAR ship detection. There are imbalances in the size distribution, azimuth distribution, and background distribution under the real data collection environment. However, since SAR ship detection is a single-class detection task, the aforementioned imbalances lack reliable descriptors during training. This paper proposes using evidence learning to obtain the epistemic uncertainty as a descriptor of biased learning on samples. Contrastive learning is used to further utilize the uncertainty label of samples to correct biased learning under intra-class imbalance. The proposed method is proven to be effective on multiple network models. AP50 reaches 94.8% on the HRSID dataset, 98.4% on SSDD dataset and 80.9% on the LS-SSDD dataset, both achieving SOTA performance.

Multi-temporal PolSAR data captures the temporal variations in polarization parameters, enabling more accurate land cover classification. Most existing multi-temporal PolSAR features rely on comparing only two-time points. These approaches can be limited in capturing cumulative changes over a longer period. To better represent the cumulative changes of land cover in the entire time-series, this paper proposes a multi-temporal PolSAR feature, namely time correlation entropy. We first extract the dominant scattering mechanism of targets from the polarization covariance matrices using matrix decomposition. Then the time correlation matrix is constructed by comparing all dominant scattering mechanism pairs in the time series. From the Shannon entropy, the entropy of the time correlation matrix, i.e., time correlation entropy, is derived to indicate the degree of changes in the land cover during the observation period. Finally, the maximum entropy principle is further applied to prove that this entropy conforms to a normal distribution. Following this corollary, a classification method based on the interval estimation of distribution parameters is proposed. We evaluate the proposed feature and classification on the salt lake classification application in Qarhan Salt Lake and Gasikule Salt Lake using Sentinel-1 images. Compared to common PolSAR features and classification methods, our method gains the best results. Besides, its results also have better regional consistency and noise resistance.

Traditional absolute radiometric calibration methods for synthetic aperture radar (SAR) face challenges in terms of flexibility, maintenance, and calibration frequency. In contrast, radiometric cross calibration can achieve rapid and timely calibration by utilizing the calibrated SAR satellites to illuminate the same ground targets. However, there are still two factors limiting the accuracy of cross calibration. First, two satellites used for cross calibration often have different incidence angles, whereas the existing methods for correcting incidence angle differences have poor performance in scene adaptation and overcorrection. Second, the stability of ground targets plays a critical role in effective cross calibration, but in practice, not all targets possess the same stability. To address the first issue, this article proposes a novel scene-driven incidence angle difference correction method. It leverages the historical information about the target scenes to determine the evaluation threshold for data blocks. Moreover, it incorporates the adaptive exponential cosine model to correct the scattering variations caused by the difference in incidence angle. To address the second issue, an uncertainty analysis method is employed to calculate the uncertainty of each data block. Then, these uncertainties are utilized to calculate weight coefficients, and the calibration constant is determined using a weighted least squares (WLS) model. Cross-calibration experimental results on the Sentinel-1A/B demonstrate an average reduction of 21.6% in the relative calibration error and 18.6% in the root-mean-square error (RMSE) compared with the traditional method, validating the effectiveness of the proposed method.

In existing superpixel-wise saliency detection algorithms, superpixel generation often is an isolated preprocessing step. The performance of saliency maps is determined by the accuracy of superpixels to a certain extent. However, it is still a challenge to develop a stable superpixel generation method. In this article, we attempt to incorporate the superpixel generation and saliency calculation steps into an end-to-end trainable deep network. First, we employ a recently proposed differentiable superpixel generation method to over-segment the synthetic aperture radar (SAR) images, which outputs the possibility that the pixels assigned to neighbor superpixels (soft superpixel). In saliency calculation part, as one of our main contributions, we propose a differentiable and computationally simple saliency model, i.e., lacunarity cue. It is inspired by the fact that generally the backscattering intensity of regions of interest (ROIs) in SAR images irregularly fluctuates, while the areas with consistent pixels are often ignored as the clusters. We improve the pixelwise box differential dimension algorithm to measure the irregularity of scattering points in a superpixel. The superpixel generation and saliency calculation can be implemented under a unified deep network. Hence, the shapes of the superpixels can be iteratively adjusted according to the saliency maps until the ROIs are correctly detected. Experiments on real SAR images with different sizes and scenes show that the saliency maps can effectively highlight the target areas, thus outperforming the state-of-the-art saliency detection models.

Ship detection from synthetic aperture radar (SAR) remote sensing images is essential for monitoring water traffic and marine safety. Numerous methods for ship detection have been developed; however, the detection of small ships presents unique challenges. SAR image characteristics, such as the sidelobe effect and blurred outline induced by the special imaging mechanism, as well as the small ship size, are the primary factors that lower the detection accuracy. This paper provides a sidelobe-aware small ship detection network for synthetic aperture radar imagery. First, considering the sidelobe effect and blurred outline, dual-pooling, i.e., average pooling and max pooling, was utilized to build a feature extraction module that lowered the effects of strong scattering points outside of the ship body and enhanced the ship body information. Second, as the bipartition process of the average pooling and maximum pooling caused some loss of original data information, different feature maps in the network were concatenated to construct a new network structure to compensate for the information lost and enrich the small ship features. Third, because the traditional loss function based on centroid distance and aspect ratio may result in the same loss function value for different prediction box sizes, a novel loss function based on the dual Euclidean distances of the corner point coordinates between the prediction box and the real box was proposed, which could accurately describe various overlapping box situations. Experiments using the Large-Scale SAR Ship Detection Dataset (LS-SSDD), SAR Ship Detection Dataset (SSDD), and AIR-SARShip dataset validated the efficacy and state-of-the-art performance.

Crop classification is one of the most important applications of polarimetric SAR images. Time-series polarimetric SAR images have the characteristics of reflecting the changes of various scattering characteristics of crops in different growth periods. However, since time-series polarimetric SAR needs to combine multiple single polarimetric SAR images, the redundancy between features is multiplied. In this paper, aiming at the problem of feature redundancy, the method of similarity measurement is used to select features from two dimensions of space and time respectively to reduce feature redundancy. Since the sample size of SAR feature images applied in supervised classification is small, it's not suitable for multiple downsampling in CNN, and a suitable classifier based on Transformer is designed. Preliminary experiments on the full polarimetric data verified the effectiveness of the proposed method.

Polarimetric features extracted from the polarimetric synthetic aperture radar data contain a wealth of target scattering information, but usually lead to the problems, such as network learning burden and high computational consumption. A multichannel fusion convolutional neural network based on scattering mechanisms was presented in this letter. First, the polarimetric features were divided into three categories according to their corresponding scattering mechanisms, and put into three network channels, respectively. Second, a new feature output was constructed based on the fusion of three-channel output features. Third, the four output features were cascaded through two fully connected layers and the Softmax classifier to get the classification result. Moreover, a new loss function was defined, combining cross entropy and average cross entropy to prevent network overfitting. Experimental results on airborne synthetic aperture radar (AIRSAR) and GF-3 data set verified the effectiveness of the proposed method in the aspect of classification accuracy and small sample.

The land cover classification has played an important role in remote sensing applications. However, most classification methods were designed based on the pixel features or local spatial features of the remote sensing image, which limits the classification accuracy and generalization. In order to further utilize the spatial information, this letter proposes a dual-branch neural network (NN) inspired by the conditional random field (CRF) model, namely CRF-Net, which takes into account the global spatial features of the image, i.e., geographic latitude-longitude information. First, a dual-branch NN is designed to extract the pixel features and coordinate features. Then, the two kinds of features are fused to realize the remote sensing imagery classification. In the experiments, randomly selected samples and spatial-disjoint samples are employed to verify the effectiveness of the proposed method for hyperspectral image (HSI) and polarimetric synthetic aperture radar (PolSAR) image classification. The experimental results show that the proposed method is superior to the traditional supervised classification methods under the spatial-disjoint sampling strategy, and can achieve the same level of accuracy under the random sampling condition.

Convolutional neural networks (CNNs) have demonstrated impressive ability to achieve promising results in PolSAR image classification. However, the traditional CNN performs convolution on local square regions with fixed sizes. The selection of these local square regions (patches) cannot fully take advantage of the boundary information of land covers and cannot search optimal neighborhoods in the whole image. To overcome these shortcomings, we propose a superpixel-based graph convolutional network (SP-GCN) for PolSAR image classification. SP-GCN utilizes superpixels as graph nodes, which makes full use of boundary information of superpixels and significantly reduces the computational cost of GCN, making it possible to apply GCN to large-scale PolSAR image classification. To reduce the impact of superpixel scale on classification results, we further propose a multiscale superpixel-based graph convolutional network (MSSP-GCN) based on the SP-GCN. Experimental results on three PolSAR datasets firmly demonstrate the superiority of the proposed SP-GCN and MSSP-GCN to other state-of-the-art methods.

In existing superpixel-wise segmentation algorithms, superpixel generation most often is an isolated preprocessing step. The segmentation performance is determined to a certain extent by the accuracy of superpixels. However, it is still a challenge to develop a stable superpixel generation method. In this article, we attempt to incorporate the superpixel generation and merging steps into an end-to-end trainable deep network. First, we employ a recently proposed differentiable superpixel generation method to over-segment the single-polarization synthetic aperture radar (SAR) image. It outputs the statistical likelihood that each pixel belongs to different superpixels. In superpixel merging part, as one of our main contributions, we propose a superpixel-wise statistical dissimilarity measure method for converting the soft superpixels set into a self-connected weighted graph. More importantly, inspired by the concept of the number of walks in graph theory, we define the$k$-order connectivity of each vertex. This definition can intelligently indicate the potential soft cluster centers and class assignments in graph. This merging method is differentiable, computationally simple, and free of empirical parameters. The superpixel generation and merging phases can be implemented under a unified deep network. The benefit is that our method can iteratively adjust the shapes of the superpixels according to the boundaries and segmentation results during training, until the satisfactory segmentation results are captured. Experimental results on real SAR images demonstrate that the segmentation precision of our proposed method is superior to other state-of-the-art methods in terms of precision and computational efficiency.

Since the number of superpixels is lower than that of pixels, superpixels can substantially speed up subsequent processing steps and have been widely used in synthetic aperture radar (SAR) image segmentation. However, in most of the existing superpixel-wise segmentation algorithms, superpixel prediction is an isolated preprocessing step and is independent of the segmentation task. The performance of the segmentation results is determined by the accuracy of superpixels. Once superpixels are generated, their shape cannot be changed in the following segmentation stage, even if the same superpixels contain pixels of different landcovers. To address this, we propose an end-to-end trainable superpixel-wise segmentation method for single-polarization SAR images. First, we design a differentiable boundary-ware clustering method for estimating task-specific superpixels. Instead of the hard association between pixels and superpixels in the existing superpixel algorithms, this method introduces the soft association map to make the clustering differentiable. Hence, it can be implemented using a simple deep fully convolutional network. In the segmentation part, we propose a novel soft graph convolution network (Soft-GCN), which takes the association map as input and performs superpixel-wise segmentation. The advantage of our method is that superpixel generation and graph convolution parts can be trained under a unified framework, until two parts obtain the optimum parameters. In the training process, it can adaptively adjust the shape of the superpixels according to the segmentation results, ensuring the superpixels correctly adhere the boundaries. Experimental results with simulated and real SAR images demonstrate that our method outperforms other state-of-the-art segmentation algorithms, while also being faster.

Currently, deep neural networks have been widely used in the field of SAR target recognition. Many researchers found that deep neural networks have an ability of denoising. In many cases, there is no need to denoise in pre-process. But the denoising ability of deep neural networks can take place of conventional denoising algorithm or not is doubtful. In this article, we explore the effect of image denoising algorithms to SAR target recognition methods based on deep neural networks. Firstly, seven traditional denoising algorithms are selected to process two SAR datasets. And these data are utilized to train two kinds of deep neural networks. After comparing and analyzing the training processes and results, we find that 1) The effect of denoising algorithms is influenced by architectures of neural networks and quality of datasets. It is difficult to find a SAR image denoising algorithm, which can improve the accuracy of any recognition network. Sometimes they even drag down the performance of recognition networks. 2) The deep networks with more layers will have better denoising ability, so the effect of denoising algorithms will decrease. For ResNet, there is no need to add the denoising processing.

Small vessels in synthetic aperture radar (SAR) images usually have weak scattering intensity and occupy only a few numbers of image pixels, resulting in a high miss detection rate during the detection process. Regarding the problem, two solutions were presented in this paper. Firstly, dual-polarimetric SAR data were used and dual-polarimetric features were adaptively fused. Comparing to single-polarization and conventional non-adaptive fusion method, it optimally enhanced the characteristics of small vessels. Secondly, the conventional feature pyramid network (FPN) was enhanced by reducing the downsampling factor, adding spatial attention, and channel attention. The added spatial attention enhanced the significant features of small vessels on the large-scale feature map; the added channel attention filtered out the spliced features maps that were benefiting small vessel detection and reduced feature redundancy. Experimental results on the small vessel data set of Sentinel-1 verified that it not only reduced the miss detection rate but also improved calculation efficiency.

Polarimetric synthetic aperture radar (PolSAR) image classification is an important part of SAR data interpretation and provides more intuitive and detailed SAR polarization information. To bridge the PolSAR data and applications, it is necessary to design a comprehensive PolSAR classification framework to achieve satisfactory results. The deep neural network (DNN) appears to be a solution for the classification issue, in which it outperforms the classical supervised classifiers under the condition of sufficient training data. However, the volume of training data will greatly limit the effectiveness of practical applications. In this article, we try to solve the dependence issue on training data in three different ways: recurrent learning, data augmentation, and postprocessing. First, the long short-term memory (LSTM) network is introduced to achieve pixel sequence learning by taking into account the spatial and polarimetric features. Second, the random neighbor pixel-block (RNPB) method is proposed to increase the number of training samples for sequence learning. Third, the conditional random field (CRF) model is employed to further improve the classification accuracy. In the experiments, three sets of PolSAR data are used to evaluate the small sample performance of the proposed classification method. With only 0.5% labeled pixels for training, the proposed RNPB-LSTM-CRF method can approach 99% overall classification accuracy for all the data sets. Compared with the existing methods, the proposed method can achieve state-of-the-art results for PolSAR image classification under the condition of 1% training samples.

Many deep learning algorithms have been successful applied for synthetic aperture radar automatic target recognition (SAR-ATR), but high recognition accuracy usually relies on large amount of labeled training data. In addition, SAR is active imaging sensor and target characteristics are quite different with varying look angles, which also reduces recognition accuracy. Multi-task learning can improve the performance of main task by learning and sharing useful information from auxiliary tasks. Based on multi-task learning, this paper fully exploits the potential of available SAR data for target classification. Two auxiliary tasks, separating target from shadow and estimating target aspect angle, are designed to obtain auxiliary information and improve the classification accuracy. The MSTAR data set proves the effectiveness of the method, and the results show that the method has good recognition accuracy.

Simultaneous multitask processing is a common requirement in synthetic aperture radar (SAR) automatic target recognition (ATR), e.g., not only the category of the target but also the aspect angle of the target need to be identified at the same time. Moreover, the target recognition network is always expected to have the capability of incremental learning, i.e., acquire the processing capabilities for new tasks while maintaining the processing capabilities for old tasks. In this paper, an incremental multitask learning method based on structured pruning is proposed. The structured pruning, originally proposed for network compression, is used to learn with dominant neuron and release parameter space of convolutional neural network for new tasks. Through iterative pruning and training of new tasks, multitask target recognition is realized in a single convolutional neural network and could simultaneously output recognition results of multiple tasks. The experiments on the MSTAR dataset show that our method can simultaneously recognize the category and aspect angle of target, while does not decrease the corresponding accuracy compared to single-task processing.

Polarimetric Synthetic Aperture Radar (PolSAR) image classification is an essential part of SAR data applications. As one of the image classification methods that can efficiently capture structural information and semantic context, the convolutional neural network (CNN) seems to be a solution for the classification issue in that it outperforms the classical supervised classifiers under the condition of sufficient training data, and it has been used in PolSAR classification widely. Simultaneously, the distance metric learning (DML) is proposed to improve the classification algorithms in performance and even in feature extraction. In this paper, DML with adaptive density discrimination regarded as a loss function, namely Magnet Loss, is applied to the classification of PolSAR images, and k-means++ is realized the clustering process for each category of training samples. Then, different classifiers are executed to replace the softmax function to achieve more accurate classification. Finally, a series of experiments are implemented to prove the effectiveness of the proposed method. Simultaneously, the samples of the coarse label are given and used to analyze the fine-grained classification algorithm by clustering.

Automatic target recognition (ATR) has made great progress with the development of deep learning. However, the target feature in synthetic aperture radar (SAR) image is not consistent with human vision, and the SAR training samples are always limited. These hard issues pose new challenges to the SAR ATR based on convolutional neural network (CNN). In this letter, we propose an improved CNN model to solve the limited sample issue via the feature augmentation and ensemble learning strategies. Normally, the high-level features that are more comprehensive and discriminative than the middle-level and low-level features are always employed for category discrimination. In order to make up the insufficient training features in the limited sample case, the cascaded features from optimally selected convolutional layers are concatenated to provide more comprehensive representation for the recognition. To take full advantage of these cascaded features, the ensemble learning-based classifier, namely, the AdaBoost rotation forest (RoF), is introduced to replace the original softmax layer to realize a more accurate limited sample recognition. Through the AdaBoost RoF method, not only are these features further enhanced by the rotation matrix but also a strong classifier is constructed by several weak classifiers with different adjusted weights. The experimental results on MSTAR data set show that the cascaded features and ensemble weak classifiers can fully exploit effective information in limited samples. Compared with the existing CNN method, the proposed method can improve the recognition accuracy by about 20% under the condition of ten training samples per class.

Scene text detection is one of the most challenging problems in computer vision and has attracted great interest. Different from generic object detection, scene text detection mainly suffers from the large variance of scale, aspect ratio, and orientation in scene text. In this paper, we propose an effective and efficient model (SEG-FPN) for scene text detection, which is based on Feature Pyramid Network (FPN) and Linking Segments (SegLink). We incorporate feature pyramid mechanism with Single Shot Detector (SSD) framework to deal with different scale texts, and link locally detectable elements to detect texts of different orientations and aspect ratios. Moreover, compared with SSD, we enlarge the feature map of deep layers to better localize the large texts and recognize the small texts accurately. Experiments on ICDAR2015 and ICDAR2013 datasets demonstrate that our method can achieve comparable performance in terms of both accuracy and time. Specifically, SEG-FPN achieves an f-measure of 0.820 at 10.3 fps for 1280*768 ICDAR 2015 Incidental text images, and an f-measure of 0.879 at 19.2 fps for 512*512 ICDAR 2013 focused scene text images.

Chinese scene text reading is one of the most challenging problems in computer vision and has attracted great interest. Different from English text, Chinese has more than 6000 commonly used characters and Chinese characters can be arranged in various layouts with numerous fonts. The Chinese signboards in street view are a good choice for Chinese scene text images since they have different backgrounds, fonts and layouts. We organized a competition called ICDAR2019-ReCTS, which mainly focuses on reading Chinese text on signboard. This report presents the final results of the competition. A large-scale dataset of 25,000 annotated signboard images, in which all the text lines and characters are annotated with locations and transcriptions, were released. Four tasks, namely character recognition, text line recognition, text line detection and end-to-end recognition were set up. Besides, considering the Chinese text ambiguity issue, we proposed a multi ground truth (multi-GT) evaluation method to make evaluation fairer. The competition started on March 1, 2019 and ended on April 30, 2019. 262 submissions from 46 teams are received. Most of the participants come from universities, research institutes, and tech companies in China. There are also some participants from the United States, Australia, Singapore, and Korea. 21 teams submit results for Task 1, 23 teams submit results for Task 2, 24 teams submit results for Task 3, and 13 teams submit results for Task 4. The official website for the competition is http://rrc.cvc.uab.es/?ch=12.

Generative adversarial network (GAN) is an artificial neural network based on unsupervised learning method. Due to its powerful model representation capabilities, GAN has been introduced to synthesize synthetic aperture radar (SAR) image data, for the real sample is difficult to acquire. Large-scale, high-resolution SAR images play an important role in promoting SAR applications, such as automatic target recognition and image interpretation. However, on account of the difficult training problem of GAN network, especially for SAR images with speckle noise, it is difficult to obtain high-resolution SAR images by simply transfer the net from optical image. Recent studies in other image fields have shown that hierarchical structure is an effective and useful way to decompose a generation task into several smaller subtasks. How to obtain more high-resolution SAR images from limited original samples through GAN is the target of our research. Therefore, in this paper, we introduce a hierarchical GAN network model to generate SAR images, through the multi-stage network, gradually improve the quality of the generated image, and finally obtain high-resolution images. The type and aspect of generated images are determined by the input of condition vectors in the last two stages. In addition, we introduce the triple loss, in which the background loss is used to imitating background clutter noise of SAR image, the condition loss is to make the generated images' type and aspect become controllable, and the global loss for getting higher image generation quality. The generated images show high similarity with the real samples.

In this paper, we improve the Siamese Networks for SAR target few-shot learning. SAR target recognition is an important branch of SAR application. It can efficiently extract target category information from complex SAR images and help humans quickly understand SAR images. However, many successful machine learning methods require large amounts of annotated data. So, few-shot learning is always a topical challenge for machine learning. We apply Siamese Networks to SAR target recognition with limited data and improved it. Our model consists of CNN encoder, similarity discriminator and classifier. Relevantly, it has two inputs and three outputs. CNN encoder is constrained by similarity discriminator and classifier. Furthermore, the larger difference from the Siamese Network is that the target category is outputted by the classifier, not by the similarity discriminator. Our method not only makes use of the advantage of metric learning to improve the accuracy of SAR target recognition with limited data, but also significantly reduces the prediction time consumption for the model based on metric learning. In the ten categories military vehicle classification task, there are only five samples for each category and a total of 2425 testing samples. Our method outperforms A-ConvNet and Siamese Networks by 15.8% and 8.41%. The prediction time consumption of Siamese Networks is 114.832s, while that of our method is 1.172s.

The characterization and understanding of microwave remote sensing image quality is essential to the monitoring performance of sensors and the proper usage of the acquired image. Point targets (e.g., passive corner reflectors and active transponders) have been widely used for microwave image resolution analysis. However, the analysis results based on point targets do not include the effects of speckle and thermal noise that are rare for point targets but common for distributed targets. Since distributed targets are common in remote sensing imaging scenes and their distinguishability is of interest in practice, the concept of an optical bar pattern target was extended to the microwave band, and a microwave bar pattern target was designed and permanently built at the National Calibration and Validation Site for High-Resolution Remote Sensors. In this letter, the main design idea is first introduced. Different backscatter coefficients of the bars were achieved for different surface roughness made by black rough gravel and a white smooth concrete plate, where the gravel size was designed per the Rayleigh roughness criterion. The experimental results using C-band airborne SAR and X-band KOMPSAT-5 SAR images are presented. A quantitative analysis shows that this target could roughly evaluate the image resolution of high-resolution microwave imaging sensor and would be a good complement to point targets.

The assessment of Synthetic Aperture Radar (SAR) image resolution is essential to characterize and improve sensor performance, and to make better application of the acquired SAR data. Trihedral corner reflectors and active transponders have been widely used as standard point targets for SAR image spatial resolution assessment. However, these point targets have limitations in straight-forward result reveal and tolerance in deployment and processing errors. In light of the bar-pattern target which widely used for optical image resolution assessment, and to assess the image resolution of the SAR sensors operating at different frequency, different platform (airborne and spaceborne), a permanent bar-pattern target was designed and realized by black gravel and greyish white concrete bars. Gravel size, bar direction and width were carefully calculated according to the requirement of long-term operation. The effectiveness of the target was preliminarily validated by C-band airborne SAR, X-band spaceborne SAR data and optical image, and the result shows that the target is suitable for the spatial resolution assessment of both high-resolution SAR and optical sensors.

Aim at the ill-posedness of vegetation biophysical variables inversion problems, the paper presents a multi-scale, multistage (MSMS) inversion approach based on field data, multi-resolution remotely sensed observations and spatial knowledge for estimating crop leaf area index (LAI). The proposed MSMS inversion method takes advantage of multiple stages inversion strategy and prior information. Firstly, Hyperion data (30 meter) is upscaled to 300 meters and 3000 meters for establishing a multi-scale data series. Secondly, a multiple scale inversion frame is constructed to update the prior knowledge by using coarse scale inversion results as the prior information for middle scale inversion process. Thirdly, the spatial information, extracted by Taylor expansion method, is applied to reduce the influence of spatial heterogeneity on LAI retrieval. At last, a multiple stage inversion process is established based on uncertainty and sensibility matrix (USM) to realize the reasonable distribution of limited remote sensing observation in the model inversion, with which the most uncertainty parameters will be retrieved from the most sensibility remote sensing data. The experiment results indicate that the methodology proposed in this paper is reasonable and accurate for LAI estimation.

Trihedral corner reflectors have been widely used as standard point targets for synthetic aperture radar (SAR) calibration. The RCS accuracy of the trihedral corner reflector is vital especially for future high-precision SAR radiometric calibration. In order to reduce the interaction between corner reflector and the ground on which the reflector deployed, and also reduce the edge diffraction through minimizing the panel external edge length, improved trihedral corner reflector was developed and it had smaller edge length than the trihedral corner reflector with the polygonous shape for a given panel area. General expression for the panel area and panel external edge length of arbitrarily self-illuminating corner reflectors was presented by parameter equation firstly. Then the edge of reflector panel was assumed as circular arc and through a numerical approximation approach, improved corner reflector of circular arc panel geometry with smaller edge length was obtained.

As a novel imaging theoretical, compressive sensing (CS) hyperspectral imaging utilizes the sparse property of the earth objects to efficiently obtain the hyperspectral cube with much less data volume. The construction of sparse basis is of great importance for CS hyperspectral imaging. In this paper, a spectral sparse basis construction method based on earth object's spectral library and redundant dictionary training is proposed. Compared with traditional DCT and wavelet basis, the sparse basis constructed by our method performs much better in simulation experiments.

Retrieving forest parameters, such as forest height, biomass from polarimetric and interferometric SAR (Pol-InSAR) images has been investigated and well demonstrated via airborne experiments for different types of forest. Terrain slope is a factor that always prevents the wide application of SAR technique due to its slant-looking imaging geometry. It induces changes of backscattering intensity, polarimetric response, etc. Regarding Pol-InSAR forest height retrieval, the effects of terrain slope on retrieval accuracy have been analyzed. The relation between terrain slope and forest height retrieval bias was established through theoretical analysis and simulation procedures based on the PolSARpro software. Based on these results, a simple topographic correction method for forest height retrieval from Pol-InSAR images was presented. This method could alleviate forest retrieval error in rugged areas easily.

Microwave staring correlation (MSC) imaging is a new type of active high-resolution microwave imaging technique. Image quality assessment is of vital for developing and monitoring any remote imaging system. This paper presented the image quality analysis for this imaging technique by theoretical analysis and simulation. MSC imaging method was introduced firstly. Then, image properties were analyzed and compared with SAR image. Speckle noise effect and side lobe of point target effect, which are intrinsic properties of SAR image, do not exist in MSC image. The quality metrics of MSC image were presented. IRW-Staring time ratio was proposed to describe the image property that image resolution improves with the number of received signals used in the image reconstruction procedure. Finally, the effects of different system parameters on the image quality were investigated via simulation experiments. The analysis results could be helpful for future imaging algorithm development and system design.