Cross-modal remote-sensing image–text retrieval (CMRSITR) is a challenging task that aims to retrieve target remote-sensing (RS) images based on textual descriptions. However, the modal gap between texts and RS images poses a significant challenge. RS images comprise multiple targets and complex backgrounds, necessitating the mining of both global and local information (GaLR) for effective CMRSITR. Existing approaches primarily focus on local image features while disregarding the local features of the text and their correspondence. These methods typically fuse global and local image features and align them with global text features. However, they struggle to eliminate the influence of cluttered backgrounds and may overlook crucial targets. To address these limitations, we propose a novel framework for CMRSITR based on a transformer architecture, which leverages global–local information soft alignment (GLISA) to enhance retrieval performance. Our framework incorporates a global image extraction module, which captures the global semantic features of image–text pairs and effectively represents the relationships among multiple targets in RS images. In addition, we introduce an adaptive local information extraction (ALIE) module that adaptively mines discriminative local clues from both RS images and texts, aligning the corresponding fine-grained information. To mitigate semantic ambiguities during the alignment of local features, we design a local information soft-alignment (LISA) module. In comparative evaluations using two public CMRSITR datasets, our proposed method achieves state-of-the-art results, surpassing not only traditional cross-modal retrieval methods by a substantial margin but also other contrastive language-image pretraining (CLIP)-based methods.

Cross-modality vessel re-identification (ReID) presents a formidable challenge in the domain of maritime surveillance, necessitating the development of robust methodologies to accurately match vessels across disparate imaging modalities. This paper introduces a novel Cross-modality Alignment Decomposition Network (CAD-Net) to address the inherent complexities associated with this task. CAD-Net incorporates a geometric-semantic cross-modal alignment module for effectively mitigating geometric and modality variances within the global features. Additionally, it integrates an adaptive local decomposition module associated with a diversity regularization, enabling the capture of local vessel features, all while circumventing the reliance on predefined part separation criteria. To address the scarcity of cross-modal vessel datasets, which are predominantly biased towards visible light modality, and to evaluate the performance of the proposed framework, we have constructed a novel dataset named KongTong-boat (KT-boat). It comprises 2,826 high-resolution images, including 1,443 RGB images and 1,383 IR images, featuring 117 distinct vessels. This dataset can be served as a new fundamental benchmark for evaluating the efficacy of cross-modality vessel ReID algorithms, filling a critical gap in the field. The experimental results obtained on the KT-boat dataset unequivocally demonstrate the remarkable effectiveness of CAD-Net in the context of cross-modality ReID. Notably, when compared to state-of-the-art cross-modality ReID algorithms applied to general cross-modality pedestrian benchmarks on KT-boat and RegDB dataset, CAD-Net consistently outperforms them across key evaluation metrics, including the rank-1 index and mean Average Precision (mAP).

Visible-infrared image re-identification (VIReID) aims to match objects with the same identity appearing across different modalities. Given the significant differences between visible and infrared images, VIReID poses a formidable challenge. Most existing methods focus on extracting modality-shared features while ignore modality-specific features, which often also contain crucial important discriminative information. In addition, high-level semantic information of the objects, such as shape and appearance, is also crucial for the VIReID task. To further enhance the retrieval performance, we propose a novel one-stage CLIP-based Modality Compensation (CLIP-MC) method for the VIReID task. Our method introduces a new prompt learning paradigm that leverages the semantic understanding capabilities of CLIP to recover missing modality information. CLIP-MC comprises three key modules: Instance Text Prompt Generation (ITPG), Modality Compensation (MC), and Modality Context Learner (MCL). Specifically, the ITPG module facilitates effective alignment and interaction between image tokens and text tokens, enhancing the text encoder's ability to capture detailed visual information from the images. This ensures that the text encoder generates fine-grained descriptions of the images. The MCL module captures the unique information of each modality and generates modality-specific context tokens, which are more flexible compared to fixed text descriptions. Guided by the modality-specific context, the text encoder discovers missing modality information from the images and produces compensated modality features. Finally, the MC module combines the original and compensated modality features to obtain complete modality features that contain more discriminative information. We conduct extensive experiments on three VIReID datasets and compare the performance of our method with other existing approaches to demonstrate its effectiveness and superiority.

Deep learning has emerged as the dominant paradigm for Synthetic Aperture Radar (SAR) automatic target recognition (ATR). However, existing learning algorithms focus primarily on single-granularity categorical recognition task without fully exploiting the high-order correlations among both data and label domains. This paper presents a multi-granularity SAR ATR framework called class hierarchy aware contrastive feature learning. The framework consists of three modules designed to extract higher-order discriminative features, enforce hierarchical feature distribution, and enable hierarchically consistent classification. The feature extractor module extracts higher-order class-specific features from two groups of multi-view samples, breaking the assumption of i.i.d. data. Furthermore, we introduce a structured contrastive regularization module to enforce a hierarchical distribution in the feature domain, thereby improving the discriminative power of the learned representations and addressing the issue of hierarchical violation. Additionally, an effective multi-granularity label representation scheme and a hierarchical classifier are developed to leverage the hierarchical relationships among classes. The framework contributes to advancing SAR ATR by providing insights into hierarchical category-aware feature learning and classification, and improving the generalization ability in scenarios with limited training samples. Experimental evaluations conducted on a benchmark SAR dataset validate the effectiveness of the proposed framework. Particularly in scenarios with limited training samples and zero-shot ATR, our framework outperforms state-of-the-art methods, achieving better performance.

In order to address the problem of Unmanned Aerial Vehicles (UAVs) being difficult to locate in environments without Global Navigation Satellite System (GNSS) signals or with weak signals, this paper proposes a localization method for UAV aerial images based on semantic topological feature matching. Unlike traditional scene matching methods that rely on image-to-image matching technology, this approach uses semantic segmentation and the extraction of image topology feature vectors to represent images as patterns containing semantic visual references and the relative topological positions between these visual references. The feature vector satisfies scale and rotation invariance requirements, employs a similarity measurement based on Euclidean distance for matching and positioning between the target image and the benchmark map database, and validates the proposed method through simulation experiments. This method reduces the impact of changes in scale and direction on the image matching accuracy, improves the accuracy and robustness of matching, and significantly reduces the storage requirements for the benchmark map database.

The idea of spatial correlation has been used in polarimetric synthetic aperture radar (PolSAR) classification for many years. It is common that the bigger the spatial correlation, the more information it contains. Though the recent advances in deep learning for PolSAR classification have achieved remarkable progress, the number of training samples is still a problem we have to face. Aiming to solve this problem, it is valuable to explore a greater spatial correlation of terrains as a priori for classification. Considering the correlated regions of the training sample as a prior, rather than a small neighborhood, the paper proposed Wishart locally constrained expansion algorithm. Based on this, a PolSAR classification algorithm is designed. The whole proposed PolSAR classification algorithm includes 3 parts: Wishart locally constrained expansion algorithm, convolution neural network training algorithm, and Markov random field post-processing algorithm. Supervised by the cluster knowledge from Wishart classifier, Wishart locally constrained expansion algorithm expands samples iteratively from the regions related to the training samples with a region-expanding technique, where very few labeled samples turn to a larger training set. Convolution neural network algorithm is designed to train convolution neural network with 3 types of training samples for a more accurate result. Convolution neural network is trained firstly with the numerous training samples generated through Wishart locally constrained expansion algorithm, and then the samples from consistency extraction and raw samples are involved to fine-tune the convolution neural network to get the improved result. Finally, Markov random field prior is used to smooth the result. Several benchmark datasets are adopted to evaluate the effectiveness of the proposed algorithm. The experimental results show that the new semi-supervised classification algorithm outperforms the state-of-the-art semi-supervised classification algorithms and classical supervised classification algorithms.