This study explores the feasibility of cross-satellite SAR ship re-identification (SAR Ship ReID). In order to overcome the adverse effects of the significant modality differences between various satellites, this study improves the method of contrastive language-image pre-training re-identification (CLIP-ReID) to adapt the cross-satellite SAR Ship Re-ID task. Specifically, we retain the text encoder of CLIP-ReID while improve the image encoder to better adapt it for SAR images. To verify its feasibility, we designed two test experiments, which respectively use (1) high-quality images as Query and low-quality images as Gallery, and (2) low-quality images as Query and high-quality images as Gallery. The proposed method achieved almost the same high Re-ID accuracy in both experiments, demonstrating the feasibility and stability. Additionally, the performance of the proposed method also surpasses several recent methods.

Improving ship classification performance in synthetic aperture radar (SAR) imagery by transferring knowledge from the related domain is a newly emerging research topic. Existing methods follow supervised or unsupervised homogeneous transfer learning techniques with certain restrictions on the use of features (homogeneous rather than heterogeneous) and data (ignoring excavate the potential of unlabeled target domain data), which may hinder further performance improvements. To address these problems, this letter proposes a dynamic joint correlation alignment (DJ-CORAL) network to conduct semi-supervised heterogeneous domain adaptation (HDA). Specifically, DJ-CORAL firstly transforms the heterogeneous features from the source and target domains into a common subspace to eliminate the heterogeneity, then simultaneously performs classifier adaptation and joint marginal and conditional distribution alignment to facilitate the domain shift minimization. Comprehensive experiments validate the superiority of the proposed DJ-CORAL network against state-of-the-art HDA methods. The codes are available at https://github.com/BUCT-RS-ML/DJ-CORAL.

Improving ship classification performance in synthetic aperture radar (SAR) imagery by the methods based on transfer learning (TL) is a newly emerging research topic and has great potential. The existing studies merely address the problem of transfer learning from a single source domain (AIS or ORS) to the target domain (SAR) based on the homogeneous transfer learning (HoTL) which requires all domains are represented by homogeneous features with same dimensions. Our work takes a step forward and attempts to address a more meaningful and challenging problem that transfers knowledge from multiple source domains for the purpose of exploring and exploiting complementarity cross source domains to assist ship classification in the target domain. To this end, our study develops a multi-source heterogeneous transfer learning (MS-HeTL) method which liberates the restriction of utilizing the exact same features for all domains, allows each domain represented by a more appropriate feature and thus improves the ship classification performance even further. Specifically, we first propose multi-source heterogeneous feature augmentation (MS-HFA) to effectively solve the challenges brought by feature heterogeneity and fully exploit the complementarity cross domains. Then a support vector machine (SVM) classification framework is specific-designed to be incorporated with the augmented feature representations in the common space to conduct knowledge transfer cross domains. Extensive experiments on two benchmark datasets, HR-SAR and FUSAR, show that the proposed method outperforms the existing methods, and demonstrates its effectiveness and advantages. All datasets and source codes are available at https://github.com/BUCT-RS-ML/MS-HeTL-via-MS-HFA.

Sea surface and oil spill have different scattering mechanisms that can be characterized by the non-circularity of the single look complex synthetic aperture radar (SLC-SAR) imagery. Based on this understanding, this paper first designs two novel non-circularity parameters then utilizes them to enhance the SAR imagery. Preliminary experiments validate that using the non-circularity enhanced SAR (NCE-SAR) imagery can help various semantic segmentation (SS) models to further improve the dark-spot segmentation performance.

There are still many challenges to be resolved in the task of ship classification in synthetic aperture radar (SAR) images, such as limited number of labeled samples in SAR domain, large variance in the same subcategory, small variance among different subcategories, etc. Transfer metric learning (TML) has the potential to mitigate those issues in the domain of interest (target domain, TD) by leveraging knowledge/information from other related domains (source domain, SD). In this article, we proposed a novel TML method, termed as geometric transfer metric learning (GTML), which achieves discriminative information preservation (DIP), geometric structure preservation (GSP), and handles the domain shift (DS) simultaneously by integrating pairwise constraints (PC), joint distribution adaptation (JDA), and manifold regularization (MR) into a unified optimization function, aiming to make full use of their complementarity to improve SAR ship classification performance. In practice, we proposed two simple but effective optimization strategies, termed as GTML-A and GTML-R, to construct optimization function. We also proposed two solutions for two typical real-world application scenarios, that is, the task of: 1) zero-labeled sample (ZLS) and 2) scarce-labeled samples (SLS) in SAR domain. The experiments conducted on both tasks show that the proposed GTML outperforms most of state-of-the-art methods. Code is available athttps://github.com/sky-Yongjie-Xu/geometric-transfer-metric-learning.

Summary Effectively obtaining the location and direction of the ship target is an important prerequisite for maritime traffic management and marine accident rescue. Thanks to the rapid development of the target detection methods based on deep learning, this article proposed a ship target detection method for multiresolution synthetic aperture radar (SAR) images based on improved region convolution neural network (R‐CNN). According to the characteristics of ship target in the SAR images, we make several improvements such as enlarging the input, proposal optimization, database target categorization, and weight balance on the basis of the standard Faster R‐CNN. The experimental results proved that the proposed method can detect target effectively and precisely in complicated scenes of multiresolution SAR images, such as in‐shore and dense targets. It has a good potential in practical application.

Supervised learning techniques are widely used in the task of ship classification in synthetic aperture radar (SAR) images in recent years. Learning distance metrics that describe the underlying distribution between data points based on the distance metric learning (DML) methods can further improve the performance of ship classification in SAR images. Traditional supervised DML methods usually learn distance metrics based on pairwise constraints, but ignore the importance of inter-class distribution discrepancy. In this study, we propose a novel DML method named distribution discrepancy maximization metric learning (DDMML) algorithm, which maximizes the maximum mean discrepancy (MMD) between different categories in the process of learning distance metrics. We adopt a high-resolution SAR ship database for experimental evaluation. The experimental results show that the proposed method outperforms the state-of-the-art DML methods.

Ship classification in synthetic-aperture radar (SAR) images is of great significance for dealing with various marine matters. Although traditional supervised learning methods have recently achieved dramatic successes, but they are limited by the insufficient labeled training data. This letter presents a novel unsupervised domain adaptation (DA) method, termed as discriminative adaptation regularization framework-based transfer learning (D-ARTL), to address the problem in case that there is no labeled training data available at all in the SAR image domain, i.e., target domain (TD). D-ARTL improves the original ARTL by adding a novel source discriminative information preservation (SDIP) regularization term. This improvement achieves an efficient transfer of interclass discriminative ability from source domain (SD) to TD, while achieving the alignment of cross-domain distributions. Extensive experiments have verified that D-ARTL outperforms state-of-the-art methods on the task of ship classification in SAR images by transferring the automatic identification system (AIS) information.

This paper proposes a new scheme for detecting ship targets in high-resolution (HR) single-channel synthetic aperture radar (SAR) images. By using the proposed spatially enhanced pixel descriptor (SEPD) and the modified density-based spatial clustering of application with noise (M-DBSCAN), this scheme can overcome typical challenges of ship detection in HR SAR images. Specifically, the proposed SEPD maps the representation for a given pixel in an SAR image into a high-dimensional feature space by embedding spatial and intensity information of its neighborhood synchronously. It enables the spatial structure information of ship targets and textural information of the sea surface to be preserved by the SEPD feature vector, leading to a significant improvement in the separability between ship targets and sea clutter. A statistical study shows that, in SEPD feature space, a large amount of pixels belonging to sea clutter gather densely in the low-value region and are surrounded by a tiny proportion of ship targets that are distributed sparsely in the high-value region. This distribution characteristic motivates us to apply a density-based clustering approach to distinguish ship targets from the sea clutter. To overcome the weakness of original DBSCAN clustering algorithm and make it suitable for the requirements of ship detection in SAR images, we propose the method of M-DBSCAN, which introduces three critical improvements, including a new dimensionality independent distance metric, a one-class clustering strategy, and an entirely deterministic approach to border points. A novel ship detector is proposed by applying M-DBSCAN to cluster pixels that are represented by SEPD descriptor. Comprehensive experiments demonstrate that the proposed method outperforms other intensity-based clustering methods ($k$ -means and fuzzy c-means), and widely used intensity threshold-based method (constant false alarm rate detector) in most circumstances, and can effectively handle various challenging situations appearing in HR images, such as sidelobes, small/weak targets, moving targets, and so on.

In this paper, a new paradigm for synthetic aperture radar (SAR) observation of ship targets at sea is presented. We firstly utilize the scattering similarity parameters to investigate the differences of scattering mechanism between ships and sea, based on which, a novel ship detection metric is proposed. Then the distribution function of the proposed metric is investigated and modeled by utilizing the method of kernel density estimation (KDE). Based on the statistical model, an automatic constant false alarm rate (CFAR) detection scheme is implemented. The proposed metric is compared with two classic polarimetric metrics, and the experimental results conducted on C-band RADARSAT-2 polarimetric SAR (Pol- SAR) data demonstrate the feasibility of the proposed metric and corresponding approach.

A major bottleneck in limiting the application of the existing methods of ship classification in synthetic aperture radar (SAR) images is the inadequate amount of labeled data available for training a classifier. However, generating ground truth involves expensive and time-consuming ground campaigns or is costly, since a high number of SAR image acquisition will be necessary. In contrast, an automatic identification system (AIS), which is an automatic tracking system used for monitoring maritime ships, can provide plenty of labeled ship samples that is relatively easier to be obtained. Inspired by these facts, this letter proposes to improve ship classification in SAR images by transferring AIS knowledge. We propose an improved multiclass adaptive support vector machine, combined with the naive geometric features (NGFs), to achieve transfer learning between the AIS domain and the SAR image domain. The experiments prove that the traditional method can be significantly improved by AIS information transfer, especially when only a few training samples in the SAR domain are available. In addition, it also shows that after feature selection, the performance of the proposed method can be close to that of the state of the art, even if by only using simpler NGFs and few training samples.

Compared with the high-resolution synthetic aperture radar (SAR) image, a moderate-resolution SAR image can offer wider swath, which is more suitable for maritime ship surveillance. Taking into account the amount of information in a moderate-resolution SAR image and the stability of feature extraction, we propose naive geometric features (NGFs) for ship classification. In contrast to the strictly defined geometric features (SGFs), the extraction of NGFs is very simpler and efficient. And more importantly, the NGFs are enough to reveal the essential difference between different types of ships for classification. To fuse various NGFs with different physical properties and discriminability, the multiple kernel learning (MKL) is utilized to learn the combination weights, rather than assigning the same weight to all features as usually applied by the traditional support vector machines (SVMs). The comprehensive experiments validate that: (1) the performance of the proposed NGF-combined MKL outperforms that of NGF-combined SVM by 3.4% and is very close to that obtained by SGF-combined MKL and (2) in terms of classifying ships in a moderate-resolution SAR image, NGFs are more feasible than scattering features.

In this paper, we address the problem of ship detection in PolSAR image. We firstly investigate the differences of scattering mechanism between ship targets and the sea surface based on the polarimetric similarity analysis. It is shown that, the sea surface scattering is dominated by the odd bounce (denoted as r1), while the scattering of ship targets are both dominated by the even bounce scattering (r2) which has been widely accepted, and the line bounce scattering (r4) which is a new find from the experiments. Based on those differences, a metric (r2+r4)/r1can be applied to distinguish ship targets from the sea surface. To further suppress the effects of sidelobes and imaging artifacts which have the same/similar scattering behaviors with ship targets, we introduce in the third eigenvalue λ3) of the coherency matrix and obtain the metric (r2+ r4)λ3/r1. The preliminary results show that a constant false alarm rate (CFAR) ship detector based on the proposed metric can obtain promising ship detection performance.

We address the problem of sea surface distribution modeling in a synthetic aperture radar (SAR) image by developing an innovative nonparametric method to tackle the main weakness of the traditional Parzen window kernel method, i.e., relatively low computation speed. We derive an explicit analytical solution of modeling sea surface distribution by a composite cubic Bézier curve and propose an adaptive segmentation strategy to improve the modeling precision. A comparative study validates that the average computation time of the proposed method is only 1/60 of the Parzen window kernel method and about 1/6 of the k-root and G0 methods. More importantly, in terms of modeling performance, the proposed method can achieve more adaptability and stability to different SAR sensors, resolutions, and sea scenes. The average goodness of fit tested on eight sea scenes of the proposed method, measured by |R̂̅2̅| (the smaller the better), is only 0.0006 and outperforms that of the Parzen window kernel method (0.0059), k-root (0.0390), and G0 (0.0678).

Selecting discriminate features and constructing an appropriate classifier are two essential factors for ship classification in a synthetic aperture radar (SAR) image. Unfortunately, these two factors are rarely considered together by existing studies. We propose a joint feature and classifier selection method by integrating the classifier selection strategy into a wrapper feature selection framework. The sequential forward floating searching algorithm is improved to conduct efficient searching for an optimal triplet of feature-scaling-classifier. Comprehensive experiments on two data sets demonstrate that the proposed method can select the optimal combination of a nonredundant complementary feature subset, appropriate scaling, and classifier to improve the performance of ship classification in a SAR image.

Probability density function (pdf) estimation of sea clutter in synthetic aperture radar (SAR) imagery has a fundamental role in constructing a constant false alarm rate (CFAR) based ship detector. This paper proposes a semi-parametric sea clutter modeling method for SAR amplitude imagery. The pdf of sea clutter is estimated point by point for each amplitude value, by selecting an optimal component from a given dictionary. For a specific point, the optimal component is selected by measuring the statistical consistency between pdfs of different components and the pdf of sample data within a local window in pdf domain. The statistical consistency is measured by Kullback-Leibler distance (KL-distance). The size of local window is determined based on smoothness criterion. Experimental results on several real SAR imageries demonstrate that the proposed method accurately models the sea clutter, and is flexible to combine with CFAR to construct a ship detector.

In this paper, we study a novel problem of classifying covert photos, whose acquisition processes are intentionally concealed from the subjects being photographed. Covert photos are often privacy invasive and, if distributed over Internet, can cause serious consequences. Automatic identification of such photos, therefore, serves as an important initial step toward further privacy protection operations. The problem is, however, very challenging due to the large semantic similarity between covert and noncovert photos, the enormous diversity in the photographing process and environment of cover photos, and the difficulty to collect an effective data set for the study. Attacking these challenges, we make three consecutive contributions. First, we collect a large data set containing 2500 covert photos, each of them is verified rigorously and carefully. Second, we conduct a user study on how humans distinguish covert photos from noncovert ones. The user study not only provides an important evaluation baseline, but also suggests fusing heterogeneous information for an automatic solution. Our third contribution is a covert photo classification algorithm that fuses various image features and visual attributes in the multiple kernel learning framework. We evaluate the proposed approach on the collected data set in comparison with other modern image classifiers. The results show that our approach achieves an average classification rate (1-EER) of 0.8940, which significantly outperforms other competitors as well as human's performance.

The advances in image acquisition techniques make recording images never easier and brings a great convenience to our daily life. It raises at the same time the issue of privacy protection in the photographs. One particular problem addressed in this paper is about covert photographs, which are taken secretly and often violate the subjects' willingness. We study the task of automatic covert photograph classification, which can be used to help inhibiting distribution of such images (e.g., Internet image filtering). By carefully collecting and investigating a large covert vs. non-covert photographs dataset, we observed that there are many features (e.g., degree of blur) that seem to be correlated with covert photographs, but counter examples always exist. In addition, we observed that image visual attributes (e.g., photo composition) play an important role in distinguishing covert photographs. These observations motivate us to fuse both low level images statistics and middle level attribute features for classifying covert images. In particular, we propose a solution using multiple kernel learning to combine 10 different image features and 31 image attributes. We evaluated thoroughly the proposed approach together with many different solutions including some state-of-the-art image classifiers. The effectiveness of the proposed solution is clearly demonstrated in the results. Furthermore, as the first study to this problem, we expect our study to motivate further research investigations.