Shikai Jiang

Ship detection in optical remote sensing imagery is pivotal for various civilian and military applications. However, due to the presence of cluttered backgrounds and the significant variations in ship scales, accurately detecting ships in complex maritime environments remains challenging. To address these challenges, we propose the Dual Attention and Scale-aware Feature Alignment Network (DASFA-Net), to effectively detect multi-scale ship targets under clutter interference. The DASFA-Net comprises two primary components: Deformable Spatial Attention Module with Channel Integration (DSAM-CI) and Bidirectional Flow Alignment Network (BiFAN). The DSAM-CI enhances discrimination between ships and distractors by jointly modeling correlations across the channel and spatial domains and applying deformable spatial attention. Concurrently, BiFAN employs a bidirectional feature flow alignment strategy to resolve spatial misalignment in feature fusion, significantly improving the quality of multi-scale feature fusion. Lastly, we design the Dynamic Alpha Complete Intersection over Union (DA-CIoU) based on prior knowledge of target characteristics to explicitly refine our detector for maritime ships. Extensive experiments on three challenging ship detection benchmarks, including HRSC2016, SCCOS, and FGSRCS have demonstrated the superiority of our DASFA.

The key to infrared small target detection lies in dis-tinguishing targets from backgrounds for accurate extraction. However, long-distance imaging with low angular resolution causes morphology loss and weak saliency, making targets easily submerged in complex backgrounds. As a result, achieving high-probability, low-false-alarm detection solely in the spatial domain is challenging. To address this, we extend to the spatiotemporal domain and propose Stream Alignment Mamba for Motion In-frared Small Target Detection (SAMamba). Specifically, we first design a Stream Mamba Block based on a dynamic state space model, which introduces bidirectional spatiotemporal scanning to explore inter-frame relationships in sequential images, not only preserving spatial correlations but also leveraging the causality of temporal cues for comprehensive global background representa-tion. Second, we develop a Cross-Frame Feature Enhancement Block to implicitly align feature streams by computing local re-sponses between neighboring and reference frames, thereby strengthening weak target features. In addition, we devise a Hi-erarchical Feature Alignment Block with adaptive adjustment strategies and position-aware mechanisms to address spatial mis-alignment during upsampling, thus enhancing the effectiveness of multi-scale target fusion and ultimately improving detection ac-curacy. Extensive experiments on IRDST, IRSatVideo-LEO, and TSIRMT datasets demonstrate that our method achieves signifi-cant advantages over state-of-the-art (SOTA) approaches.

Image stitching aims to align two images from different perspectives. For infrared remote sensing images, the low resolution, lack of strong feature points, and the presence of large texture-less regions make it difficult to achieve effective feature matching and high-quality image stitching results. In the field of remote sensing stitching, the primary challenge is how to effectively extract features, reduce the influence of parallax, and improve registration accuracy. To improve the image stitching performance and obtain parallax-tolerant fine registration results, we propose a two-stage image stitching method based on unsupervised learning. Firstly, in the first stage, we use a multi-level feature extraction network to effectively extract image correlation features, progressively refining the registration from coarse to fine, thus ensuring performance under large-baseline conditions. Secondly, by utilizing a discrete feature detection module in the multi-level network, we remove anomalous feature regions and recombine effective local feature regions, enabling the fusion of detailed features with global features and improving registration accuracy. Finally, in the second stage, an image fine adjustment module is applied to process the image background and foreground, further eliminating parallax artifacts and improving registration accuracy. Compared with existing methods, our method has advantages in both registration accuracy and parallax tolerance. Extensive experiments demonstrate that our method effectively registers and stitches infrared remote sensing images on both the self-built infrared remote sensing dataset and the publicly available UDIS-D dataset, outperforming current state-of-the-art (SoTA) methods in terms of performance. The code is available at https://github.com/Shamfeeding/Image-Stitching.

Image registration is a prerequisite for many multi-source remote sensing image fusion applications. However, due to differences in imaging factors such as sensor type, imaging time, resolution, and viewing angle, multi-source image registration faces challenges of multidimensional coupling such as radiation, scale, and directional differences. To address this issue, this paper proposes a Siamese network based on cross-domain robust feature decoupling as an image registration framework (CRS-Net), aiming to improve the robustness of multi-source image features across domains, scales, and rotations. Firstly, we design Siamese multiscale encoders and introduce a rotation-invariant convolutional layer without additional training parameters, achieving natural invariance to any rotation. Secondly, we propose a modality-independent decoder that utilizes the self-similarity of feature neighborhoods to excavate stable high-order structural information. Thirdly, we introduce cluster-aware contrastive constraints to learn discriminative and stable keypoint pairs. Finally, we design three multi-source remote sensing datasets and conduct sufficient experiments. Numerous experimental results show that our proposed method outperforms other SOTA methods and achieves more accurate registration in complex multi-source remote sensing scenes.

Anomaly detection has emerged as a critical area of research in hyperspectral imagery processing, focusing on detect sparse, small targets with spectral and spatial features deviating from the background without prior information. The approach of anomaly detection based on reconstruction differences is a leading method in deep learning for hyperspectral anomaly detection (HAD). A key challenge is the accurate estimation of complex backgrounds. The essence of this challenge lies in accurately reconstruct background regions while inferring the latent background of anomaly regions. In this article, we propose a novel method called dual-window spectral diffusion (DWSDiff) for hyperspectral anomaly detection. To address the challenge of complex background estimation in HSIs, we developed a spectral diffusion model specifically tailored for HSI. This model achieves precise background estimation through an iterative spectral diffusion and reverse reconstruction process. We also introduced a dual-window strategy to mitigate the influence of anomaly extension area within the neighborhood on background estimation. Moreover, the scarcity of paired labeled hyperspectral images from the same scene, with and without anomalies, limits the model’s ability to learn features between anomaly and background. To address the shortage, we devised an anomaly generation strategy based on principal component analysis and the linear spectral mixing model. Building on these, we designed a training and inference framework that integrates spectral diffusion, reverse background reconstruction, and target detection. Experimental results on the Airport–Beach–Urban (ABU) hyperspectral datasets demonstrate that DWSDiff outperforms 20 state-of-the-art HAD methods across 6 different AUC metrics.

Remote sensing image change detection is an important means for earth monitoring task, which has a wide application prospect. In multi-temporal optical remote sensing, there are inherent differences in factors such as lighting and sensors. This leads to the coupling of content change and image style change, making it difficult to distinguish. Therefore, a meaningful thinking for change detection is to decouple and capture the real changes of ground objects from multi-temporal images. Based on this motivation, a novel general change detection architecture is explored, StyleFormer. It first proposes the concept of spatial-temporal style base and no longer constrains to semantic representation in a single image style. Instead, it introduces a spatial-temporal interactive style projection layer between bi-temporal images, which projects the unseen diverse styles into the consistent expression space for change detection. Furthermore, an iterative interaction strategy of Transformer and CNN features is proposed to mine spatial-temporal context information more finely. It solves the lack of local perception and non-hierarchical features in ViT, and improves the model expression ability. After that, a change prior-guided cross-attention is introduced to fuse bi-temporal features. It can adaptively enhance the change feature and improve the perception ability for small changes in remote sensing scenes. Sufficient experiments on four typical change detection datasets show that the proposed method is superior to the state-of-the-art methods. Especially on the dataset CDD-CD and SYSU-CD, the F1 score improved to 96.08% and 83.29%. The code of this work will be available at https://github.com/Tom-Dongfang/change-detection-StyleFormer.

Ship identification in optical remote sensing images is essential for a wide range of civil and military applications, including maritime rescue, port management and sea area surveillance. However, current studies focus mainly on ship detection or coarse-grained ship size identification, rather than fine-grained type identification. Moreover, interference from clouds and port facilities as well as complex conditions such as occlusion and shadows increase the difficulty of ship type identification. To address these problems, we propose a novel ship identification method by employing adjacent-branched saliency filtering and prior representation-based classification strategies, which achieves high-precision type recognition performance for large and medium-sized ships under complex environmental interference conditions. In the candidate region extraction stage, a multi-scale feature aggregation structure that utilizes feature map fusion in adjacent layers and receptive field mining within the same extraction branch is presented, providing fine representation of the location and edge characteristics of ship targets in complex scenes. In the classification stage, the low-rank term describing inter-class differences and the graph-based regularization term describing intra-class differences are added to the representation model as prior constraints, which can correctly classify ships in the presence of complex environmental interference such as occlusion and shadow. Experimental results on two high-quality ship datasets indicate that the proposed method realizes state-of-the-art identification performance compared with benchmark methods.

Remote sensing image change detection is a valuable technology for analyzing the earth observation data. It has significant application value in resource monitoring, disaster assessment and urban planning. However, current change detection methods have not fully explored the interrelationships between bi-temporal data, and the extraction process of change information lacks prior guidance and constraints. Therefore, it is easy to produce missed detections and false alarms when facing complex backgrounds and variable objects in remote sensing images. To tackle such issues, we propose a cross-scale guided high-order feature interaction change detection network for dual temporal images. Specifically, a cross-scale guided dual encoderdecoder backbone is proposed to constrain the reconstruction process of change objectives, and guide geometric prior to optimize the representation of target structures. Next, an efficient high-order feature interaction module is designed, employing multi-level receptive fields to enhance the perception ability for multiscale features. Moreover, we construct a bi-temporal feature alignment fusion module, which decouples and filters out the interference of background pseudo changes through interactive perception of spatial-temporal differences. Comprehensive experimental validation is undertaken on four representative change detection datasets (LEVIR-CD, WHU-CD, DSIFN-CD and S2Looking). The findings demonstrate that the network demonstrated state-of-the-art performance.