Self-supervised contrastive learning methods offer a promising approach to the small-sample synthetic aperture radar (SAR) automatic target recognition (ATR) problem by autonomously acquiring valuable visual representations from unlabeled data. However, current self-supervised contrastive learning methods primarily generate supervisory signals through augmented views of the original images, thereby underutilizing the rich information inherent in SAR images. To overcome this limitation, we integrate SAR targets’ geometric and physical properties, as captured in SAR target segmentation semantic maps and attribute scattering center reconstruction maps into the contrastive learning stage. Moreover, we propose a novel multimodal views contrastive learning method which contains two stages. In the contrastive learning stage, we leverage a large amount of unlabeled data for both intra-modal and cross-modal contrastive learning, thereby transferring discriminative information from these two views to the original image features to learn the feature representation. In the supervised training stage, the linear classifier is trained using a small number of labeled samples to partition the feature representation space and migrate to the downstream recognition task. Experiment results demonstrate that the proposed method achieves superior recognition performance in SAR small-sample ATR tasks and exhibits robust generalization capabilities, thereby providing additional discriminative information that augments target representation.

When incorporating spatial priors (such as sparsity and low occurrence frequency) to identify anomalies, existing hyperspectral anomaly detection (HAD) methods typically operate on the original high-dimensional spectral bands or features obtained through dimensionality-reduction techniques. However, they often neglect the explicit modeling and exploitation of the internal discriminative structure inherent in bands, consequently diluting the discriminative spectral information crucial for effective anomaly-background separation. This limitation becomes particularly pronounced when processing hyperspectral data, where the rich spectral heterogeneity could provide valuable discriminative cues for HAD. To address this issue, we propose a super-band quaternion-guided sparse network (SuperB-QSNet), which explicitly encodes the spectral discriminant structure while retaining the advantage of spatial sparse priorities. First, a heterogeneous-homogeneous super-band compression (HH-SBC) organizes spectral bands into meaningful hetero-groups. It naturally reduces redundancy and retains the representative spectral information. Second, the quaternionic sparse recovery (QSR) model encodes both spatial structures and intra-group spectral correlations via hypercomplex algebra, ensuring robust anomaly preservation and enhanced detection of spectral variations. Third, we design a differences-guided quaternionic convolution detector (DG-QCD) to maintain spectral coherence via group processing and enhance discriminative capability via inter-group difference operation. Extensive experiments show that SuperB-QSNet enhances anomaly detection accuracy while ensuring computational efficiency across real-world hyperspectral datasets.

Synthetic aperture radar (SAR) automatic target recognition (ATR) is extensively applied in both military and civilian sectors. Nevertheless, test and training data distribution may differ in the open world. Therefore, SAR out-of-distribution (OOD) detection is important because it enhances the reliability and adaptability of SAR systems. However, most OOD detection models are based on maximum likelihood estimation (MLE) and overlook the impact of data uncertainty, leading to overconfidence output for both in-distribution (ID) and OOD data. To address this issue, we consider the effect of data uncertainty on prediction probabilities, treating these probabilities as random variables and modeling them using Dirichlet distribution. Building on this, we propose an evidential uncertainty aware mean squared error (UMSE) loss function to guide the model in learning highly distinguishable output between ID and OOD data. Furthermore, to comprehensively evaluate OOD detection performance, we have compiled and organized some publicly available data and constructed a new SAR OOD detection dataset named SAR-OOD. Experimental results on SAR-OOD demonstrate that the UMSE approach achieves state-of-the-art (SOTA) performance. The code and data are available at:https://github.com/Xiaoyan-Zhou/UMSE-SAR-OOD-Detection.

Image matching is a primary technology to fuse the complementary information from optical and SAR images. Due to the high nonlinear radiometric and geometric relationship, the optical and SAR image matching task remains a widely unsolved challenge. In this study, we propose to use a Siamese convolutional neural network (CNN) architecture to learn pixelwise deep dense features. The proposed network is able to balance the learning of high-level semantic information and low-level fine-grained information, which is nonnegligible for feature matching task. Under the local searching framework, the loss function is defined based on the score map produced by the sum of squared differences (SSDs) between the learned pixelwise dense features of local optical and the SAR image patches, with a fast implementation in the frequency domain. The hardest negative mining strategy is adopted to increase the discrimination of the network. Extensive experiments are conducted on optical and SAR image pairs of different spatial resolution and different landcover types, verifying the superiority and robustness of the proposed method in terms of matching accuracy and matching precision.

Studying synthetic aperture radar automatic target recognition (SAR-ATR) under small samples can get rid of the sample dependence and improve the practicality of the deep learning model. However, the deep learning model trained with small samples is prone to overfitting. In order to solve the above problem of SAR-ATR, a novel loss function called limited data loss function (LDLF) is proposed in this letter, which organically combines the cross-entropy loss function and the contrastive loss function. The LDLF supervises the convolutional neural network (CNN) to learn strong generalization performance features. Then, for simplifying the training and testing of CNN based on LDLF, a feature-combined module is proposed. This module makes up for the limitation that the model input must be image pairs and simplifies the testing process of CNN. Experiments on moving and stationary target acquisition and recognition (MSTAR) datasets show that the proposed loss function is better than the cross-entropy loss function and superior to the existing methods in synthetic aperture radar (SAR) image target recognition using small samples data.

To fully explore the complementary information from optical and synthetic aperture radar (SAR) imageries, they need first to be coregistered with high accuracy. Due to the vast radiometric and geometric disparity, the problem to match high-resolution optical and SAR images is quite challenging. The present deep learning-based methods have shown advantages over the traditional approaches, but the performance increment is not significant. In this article, we explore a better network framework for high-resolution optical and SAR image matching from three aspects. First, we propose an effective multilevel feature fusion method, which helps to take advantage of both the low-level fine-grained features for precious feature location and the high-level semantic features for better discriminative ability. Second, a feature channel excitation procedure is conducted using a novel multifrequency channel attention module, which is able to make image features of different types and multiple levels effectively collaborate with each other and produce image matching features with high diversity. Third, the self-adaptive weighting loss is introduced, with which, each sample is assigned with an adaptive weighting factor, and therefore, information buried in all nearby samples can be better exploited. Under a pseudo-Siamese architecture, the proposed optical and SAR image matching network (OSMNet) is trained and tested on a large and diverse high-resolution optical and SAR dataset. Extensive experiments demonstrate that each component of the proposed deep framework helps to improve the matching accuracy. Also, the OSMNet shows overwhelming superior to the state-of-the-art handcrafted approaches on imageries of different land-cover types.

In this article, we propose a fast superpixel region merging algorithm for synthetic aperture radar (SAR) image segmentation. With our previously proposed adaptive superpixel generation approach (ALFCE), an initial over-segmentation superpixel map for SAR imagery can be obtained. A sketch edge map is used here to eliminate the mixed superpixels to refine the over-segmentation. Then, we focus on rapid superpixel merging for efficient and accurate SAR image segmentation by using the statistical region merging (SRM) framework. This article proposes a new merging order with the consideration of statistical dissimilarity measure and common boundary length penalty, as well as the homogeneity constraint for each superpixel pair. For the merging predicate, we define an adaptive merging threshold according to the image complexity, making the proposed superpixel merging no need to set any merging parameters in advance. Disjoint set is utilized in this article to map the superpixel pairs to pixel pairs for the sake of fast region merging, which has a low computation cost even with the increasing of superpixels. Experimental results on synthetic and real SAR images demonstrate that the segmentation precision of our proposed method can reach more than 85% and also superior to other state-of-the-art methods in terms of computational efficiency.

This article proposes an efficient and adaptive statistical superpixel merging approach with edge penalty for polarimetric synthetic aperture radar (PolSAR) image segmentation. Based on the initial superpixel over-segmentation result obtained by our previously proposed adaptive polarimetric superpixel generation algorithm (Pol-ASLIC), this work achieves efficient and accurate PolSAR image segmentation by merging superpixels using the statistical region merging (SRM) framework. This article proposes to define a new dissimilarity measure between superpixels, which takes the edge penalty into consideration, leading to a reasonable and accurate merging order for superpixel pairs. With regard to the merging predicate of superpixels, a polarimetric homogeneity measurement (HoM) is used to define the merging threshold, making the merging predicate and merging threshold adaptive to the PolSAR image content. Experimental results on three airborne and one spaceborne PolSAR data sets demonstrate that the proposed approach can effectively improve the computation efficiency and segmentation accuracy in comparison with state-of-the-art merging-based methods for PolSAR data. More importantly, the proposed approach is free of parameters and easy to use.

Due to the speckle noise and complex geometric distortions within SAR images, it is still a challenge to develop a stable method that can produce superpixels with both high boundary adherence and visual compactness with low computational costs at the same time. In this paper, we propose an adaptive superpixel generation approach with linear feature clustering and edge constraint for synthetic aperture radar (SAR) images, which consists of three stages. First, the local gradient ratio pattern of each pixel in SAR imagery is extracted as features, which was previously proposed by us for SAR target recognition and has been proven to be insensitive to speckle noise. Second, we propose to use the feature-ratio-based edge detector with Gauss-shaped window instead of the traditional rectangle-shaped window to obtain the edge strength map and final edges for SAR images. Finally, a modified normalized cut (Ncut)-based superpixel generation strategy is adopted using a distance metric that simultaneously measures both the feature similarity and space proximity. In this strategy, we approximate the similarity measure through a positive semidefinite kernel function rather than directly using the traditional eigen-based algorithm. Therefore, the objective functions of weighted local K-means and Ncuts can achieve the same optimum point by appropriately weighting each point in this feature space, which greatly reduces the computation cost. During the linear feature clustering, the coefficient of variation is used to automatically determine the tradeoff factor between the feature similarity and space proximity, which helps change the superpixel shape and size adaptively according to the image homogeneity. Furthermore, the edge information is also introduced to constrain the clustering for the sake of high boundary adherence. By bridging the local K-means clustering and Ncuts, as well as the benefits of edge constraint, our method not only produces superpixels with good boundary adherence but also captures the global image structure information. Experimental results with simulated and real SAR images demonstrate the effectiveness of our proposed method, which performs better than other state-of-the-art algorithms.

For synthetic aperture radar (SAR) image land cover classification, traditional feature-based methods are not always effective because of the heavy multiplicative noise. To solve this problem, we herein propose a new classification method for SAR images considering adaptive spatial contextual information. In contrast to preceding studies, the spatial contextual information of the SAR images is exploited via composite kernels (CKs). Additionally, an image superpixel strategy is employed to design an adaptive neighborhood, which enables the extraction of more accurate spatial information than a fixed-size neighborhood. Specifically, a modified superpixel map is first generated to produce the neighborhood. With this neighborhood, a context kernel is then defined by means of the Gaussian radial basis function. The resulting context kernel is combined with the conventional feature kernel via the designed CKs scheme. The relative proportion of these two kernels is controlled by a weight parameter. The label of each pixel is predicted by feeding the final CKs into a support vector machine classifier. Experiments on two real SAR images demonstrate that the proposed method can greatly improve the classification performance, both visually and quantitatively, in comparison to other traditional feature-based methods.

The classic constant false alarm rate edge detector with a rectangle-shaped filter has been proven to be effective and widely used in polarimetric synthetic aperture radar (PolSAR) images. However, in practical use, the assumption of complex Wishart distribution is often not respected, particularly in heterogeneous urban areas. In addition, as a simple smoothing filter, the rectangle-shaped window is often shown to be easy to incur false edge pixels near true edges. Therefore, its performance is limited. To overcome this restriction, we propose a new edge detector for PolSAR images, which utilizes the spherically invariant random vector product model to estimate the normalized covariance matrix for each pixel, and then replace the rectangle-shaped filter with a Gauss-shaped filter. The performance of our proposed methodology is presented and analyzed on two real PolSAR data sets, and the results show that the new edge detector attains better performance than the classic one, particularly for urban areas.

In this paper, we propose a theoretically new and effective feature for SAR image classification. The new feature combines traditional gray level co-occurrence matrix (GLCM) textural feature and the recent multilevel local pattern histogram (MLPH) feature. It can not only describe intrinsic property of land-cover/land-use surfaces, corresponding to textural information, but it also captures both local and global structural information. Experiments on real SAR images demonstrate that the proposed feature obtains better results than the original GLCM and MLPH features in SAR image classification.

The presence of multiplicative noise in synthetic aperture radar (SAR) images makes segmentation and classification difficult to handle. Although a fuzzy C-means (FCM) algorithm and its variants (e.g., the FCM_S, the fast generalized FCM, the fuzzy local information C-means, etc.) can achieve satisfactory segmentation results and are robust to Gaussian noise, uniform noise, and salt and pepper noise, they are not adaptable to SAR image speckle. This letter presents a kernel FCM algorithm with pixel intensity and location information for SAR image segmentation. We incorporate a weighted fuzzy factor into the objective function, which considers the spatial and intensity distances of all neighboring pixels simultaneously. In addition, the energy measures of SAR image wavelet decomposition are used to represent the texture information, and a kernel metric is adopted to measure the feature similarity. The weighted fuzzy factor and the kernel distance measure are both robust to speckle. Experimental results on synthetic and real SAR images demonstrate that the proposed algorithm is effective for SAR image segmentation.

Since superpixel takes spatial relationship between pixels into account, which makes the image classification process more understandable and the results more satisfactory, superpixel-based classification methods have been widely studied in recent years. However, due to speckle noise, traditional superpixel generating algorithms still have some drawbacks for synthetic aperture radar (SAR) image. In this letter, we propose a novel superpixel generating algorithm based on pixel intensity and location similarity (PILS) for SAR image. In addition, for the sake of image classification, features of Gabor filters and gray level co-occurrence matrix (GLCM) are extracted from each superpixel. The proposed superpixel generating method has the following three characteristics: (1) the terrain boundaries of SAR image are preserved well; (2) the method has more robustness against speckle noise; and (3) it has high computational efficiency. Experiments on synthetic and real SAR images demonstrate that our method significantly outperforms several state-of-the-art superpixel methods and PILS superpixel-based classification obtains better results than other pixel-based methods.

As a fast on-chip SRAM managed by software (the application and/or compiler), Scratchpad Memory (SPM) is widely used in many fields. This paper presents a Simple Scalar-based multi-level SPM memory hierarchy architecture simulator Sim-spm. We simulate the hardware of the multi-level SPM memory hierarchy successfully by extending Sim-outorder, which is an out-of-order simulator from Simple Scalar. Through the simulating memory method, the simulation framework of the multi-level SPM memory hierarchy has been built under the existing ISA (Instruction Set Architecture), which largely reduces the requirement to modify the existing compiler. The experimental results show that Sim-spm can accurately simulate the running state of the processor with a multi-level SPM memory hierarchy architecture, and it has a good prospect for the research of multi-level SPM memory hierarchy architecture.