Maritime activities are essential aspects of the human society. Accurate classification of ships is vital for maritime surveillance and meaningful to numerous civil and military applications. However, most studies conducted are limited to the coarse-grained ship classification. Few studies on fine-grained ship classification have been undertaken despite its accuracy and practicability. In this study, we construct a new benchmark for fine-grained ship classification which consists of 23 fine-grained categories of the ships. Besides the category label, the benchmark contains several other attribute information. To solve the problem of inter-class similarity, an attribute-guided multi-level enhanced feature representation network (AMEFRN) is proposed. Concretely, a multi-level enhanced visual feature representation is designed to fuse the reweighted regional features in order to focus more on the silent region and suppress the other regions. Further to this, considering the complementary role of attribute information in ship identification, an attribute-guided feature extraction branch is proposed which extracts the auxiliary attribute features by utilizing the attribute information as supervision. Finally, the attribute features and the enhanced visual features jointly function as feature representation for classification. Compared to other existing classification models, AMEFRN has better performance with an overall accuracy rate of 93.58% on the established fine-grained ship classification dataset. Moreover, it can be easily embedded into most CNN models as well as trained end-to-end.

Deep learning-based methods have achieved great success in target detection tasks of computer vision, but when it comes to Synthetic Aperture Radar (SAR) image ship detection, some new challenges appear because of the wide swath of images, diverse appearances of ships and lack of detail information, which make the detection inefficient and less effective. Aiming to these issues, in this paper, a lightweight feature optimizing network (LFO-Net) based on popular single shot detector (SSD) model is proposed for single polarization SAR image ship detection. Firstly, a simpler structure called lightweight single shot detector (LSSD) is designed, which can be trained from scratch and can reduce the training and testing time without accuracy cost. Secondly, a new bi-directional feature fusion module including one semantic aggregation block and one feature reuse block is proposed to improve the performance of multi-scale targets detection by enhancing the features of both low feature layers and high feature layers. Then the features are further optimized by leveraging attention mechanism, which is beneficial to catch the silent information more efficiently. A set of experiments are implemented to verify the effectiveness of the proposed method using the public SAR ship detection dataset (SSDD). The results show that the proposed method has significant advantages in both speed and accuracy, and outperforms other state-of-art methods. Additionally, a test on GF-3 satellite SAR data with multiple modes verifies the generalization performance of this model.

With the increasing of cross-modal data, cross-modal retrieval has attracted more attention in remote sensing (RS), since it provides a more flexible and convenient way to obtain interesting information than traditional retrieval. However, existing methods cannot fully exploit the semantic information, which only focuses on the semantic consistency, and ignore the information complementarity between different modalities. In this letter, to bridge the modality gap, we propose a novel fusion-based correlation learning model (FCLM) for image-text retrieval in RS. Specifically, a cross-modal-fusion network is designed to capture the intermodality complementary information and fused feature. The fused knowledge is furtherly transferred to supervise the learning of modality-specific network by knowledge distillation, which is helpful in improving the discriminative ability of feature representation and enhancing the intermodality semantic consistency to solve the heterogeneity gap problem. Finally, extensive experiments have been conducted on a public dataset and experimental results have shown that the FCLM method is effective in performing cross-modal retrieval and outperforms several baseline methods.

Remote sensing image scene classification (RSISC) is an active task in the remote sensing community and has attracted great attention due to its wide applications. Recently, the deep convolutional neural networks (CNNs)-based methods have witnessed a remarkable breakthrough in performance of remote sensing image scene classification. However, the problem that the feature representation is not discriminative enough still exists, which is mainly caused by the characteristic of inter-class similarity and intra-class diversity. In this paper, we propose an efficient end-to-end local-global-fusion feature extraction (LGFFE) network for a more discriminative feature representation. Specifically, global and local features are extracted from channel and spatial dimensions respectively, based on a high-level feature map from deep CNNs. For the local features, a novel recurrent neural network (RNN)-based attention module is first proposed to capture the spatial layout information and context information across different regions. Gated recurrent units (GRUs) is then exploited to generate the important weight of each region by taking a sequence of features from image patches as input. A reweighed regional feature representation can be obtained by focusing on the key region. Then, the final feature representation can be acquired by fusing the local and global features. The whole process of feature extraction and feature fusion can be trained in an end-to-end manner. Finally, extensive experiments have been conducted on four public and widely used datasets and experimental results show that our method LGFFE outperforms baseline methods and achieves state-of-the-art results.

Deep learning (DL) has increasingly become the mainstream technology for hyperspectral image change detection (HSI-CD). However, these methods lack transparency and often overlook priors in hyperspectral images (HSIs), making it difficult to extract more generalizable features directly from them. Moreover, the performance of DL-based methods typically depends heavily on a large corpus of high-quality labeled data, which is often impractical and expensive in real-world scenarios, particularly for complex hyperspectral images (HSIs). To address these issues, we propose a semi-supervised deep interpretable network for HSI-CD. Specifically, by applying structured sparse prior constraints, we propose a multi-scale convolutional sparse coding (MSCSC) model to capture shared and private sparse coefficients (SSC and PSCs) across different scales, extracting multi-scale features while reducing redundancy. We then unfold the proposed MSCSC model to establish an MSCSC-guided deep interpretable network that serves as the encoder, namely MSCSCNet, in which each network module is model-driven, enhancing the transparency of internal mechanisms and extracting more fundamental features. Additionally, we devise a two-stage semi-supervised training strategy for MSCSCNet using limited labeled data, combining a change-sensitive teacher-student self-distillation (CS-TSSD) paradigm with a novel loss function to reduce the annotation dependency and remain sensitive to change components. The proposed method not only enhances the transparency of multi-scale feature extraction but also accomplishes semi-supervised learning to extract meaningful representations, effectively integrating the advantages of model-driven and data-driven approaches. Comparative experiments on benchmark datasets demonstrate the effectiveness of our method over existing approaches. Code is available at https://github.com/Jiahuiqu/MSCSCNet.