Accurate segmentation of multiple on-orbit spacecraft remains difficult in deep-space imagery because the scenes contain large uniform backgrounds, fine structural details, and limited labeled data. To address this problem, we propose SpaceSeg, a segmentation framework that adapts a vision foundation model to the spacecraft domain. The framework introduces a Multi-Scale Hierarchical Attention Refinement Decoder (MSHARD) to improve cross-scale feature decoding, a Spatial Domain Adaptation Transform (SDAT) training strategy to improve robustness to representative space-imaging disturbances, and a task-oriented objective that jointly optimizes segmentation accuracy and IoU-prediction quality. A lightweight connected-component-analysis module is also integrated into the pipeline for instance-aware target organization in multi-spacecraft scenes. We further construct SpaceES, a multi-scale on-orbit multi-spacecraft semantic segmentation dataset covering four space backgrounds and 17 spacecraft types. On SpaceES, SpaceSeg achieves 89.87% mIoU and 99.98% mAcc, setting a new state of the art among all evaluated baselines, surpassing the strongest competing method by 1.38 percentage points in mIoU with 59.6% fewer parameters, and exceeding the vanilla SAM2 baseline by 5.71 percentage points. Hardware-in-the-loop simulation and real satellite-to-satellite imagery experiments further support the practical relevance of the proposed method. Dataset and code are publicly available at https://github.com/Akibaru/SpaceSeg.

Visual 3D reconstruction is a key technology in the field of computer vision, with significant implications for tasks such as robot manipulation, autonomous driving, and virtual reality. In recent years, methods based on neural radiance fields and Gaussian splatting have gained considerable attention due to their outstanding performance. Surface mesh reconstruction based on the 2D Gaussian model achieves high accuracy, providing critical information for subsequent target-centered perception tasks. However, challenges such as self-occlusion caused by the complex structure of spacecraft, limited observation positions due to orbital constraints, and the high cost of orbital transfer restrict data acquisition. These limitations prevent comprehensive target information from being obtained, as is possible in ground-based sampling, resulting in reconstruction failures or reduced quality. To address the above problem, this paper proposes a uncertainty-aware 2D Gaussian splatting method for 3D mesh reconstruction under restricted viewpoint observation. First, the fixed color value of the 2D Gaussian ellipsoid is expanded into a probability distribution to measure uncertainty. Then, a color negative log-likelihood loss function is designed to train the Gaussian elements to estimate the mean and variance of the probability distribution. The proposed method is validated on a spacecraft reconstruction dataset collected from our local darkroom environment, demonstrating its effectiveness through qualitative and quantitative comparisons of 3D mesh reconstruction, uncertainty estimation, and novel view synthesis.

In this paper, the dual spacecraft pursuit-evasion game problem under incomplete information is investigated, and a strategy-solving method for the incomplete information pursuit-evasion game based on particle swarm optimization and unscented particle filter (PSO-UPF) estimation is proposed. The completeness of the information available about the target’s cost function, which is determined by the weighting information, has a significant impact on the success of the pursuing strategy. For the cost function is unknown in incomplete information scenarios, a research framework of the pursuit-evasion game based on following observation and one-sided pursuit two stages is established. Besides, to describe the more accurate motion of the spacecraft, a Schweighart-Sedwick (SS) dynamic model is introduced that considers the effect ofJ2perturbation. Firstly, an equilibrium strategy for the SS model-based pursuit-evasion problem is derived under complete information. Next, for the incomplete information scenarios, an estimation method based on PSO-UPF of weight matrix information is established, which allows the cost function to be determined by the estimation method in the observation stage. Then, the pursuit strategy is re-designed in the one-sided pursuit stage based on the estimated cost function. Finally, the performance of the proposed method is validated by simulation. The results demonstrate that the approach can achieve good performance by efficiently estimating the weight information in the opponent’s cost function.

Multimodal remote sensing images provide complementary information, enhancing the effectiveness of object detection tasks in open-world scenarios. To address the imbalance of information richness between modalities in multimodal object detection, we propose a simple yet effective multi-source image object detection method (DKDNet). Our contributions are twofold: (a) we introduce diffusion deformation convolution (DDConv), which combines deformation convolution with adaptive long-range receptive fields to further enhance the ability to perceive object pose variations and capture distant information. (b) We propose the bidirectional feature distillation and information complementary fusion network (BDFusion), where different modalities exchange information through a knowledge distillation strategy, explicitly enhancing the information interaction between modalities. Finally, we adaptively build spatial domain complementarity between different modalities via self-correction, revealing implicit correlations. Experimental results on the publicly available vehicle detection in aerial imagery (VEDAI) dataset and the optical and SAR ship detection dataset (OSSDD), collected in the Suez Canal region, demonstrate that our proposed method achieves superior performance with acceptable inference time, making it suitable for various realworld scenarios.

Point set registration is an essential technique in the field of machine vision. In this article, we propose a robust global optimal solution to for the point set registration of feature points extracted from visual images, used in remote (300–120 km) space target tracking and targeting tasks. Specifically, we begin with cases where correspondences among point sets are known, establishing a cost function centered on maximizing the consensus set, wherein rotational and translational parameters are determined using voting methods and the branch-and-bound (BnB) algorithm, respectively. We then adapt this foundation to tackle the more challenging scenario of unknown correspondences in simultaneous pose and correspondence registration by adjusting the cost function and BnB bounding functions, supplemented with nested iterations to accurately determine rotation and translation parameters. Finally, the comprehensive experimental comparisons executed across synthetic and real datasets, along with ground-based spacecraft pose measurement setup, illustrate that, compared to existing methods, our proposed approach achieves precise estimations under the influence of noise and outliers. Moreover, compared to the globally nested BnB scheme, our method reduces computational complexity and enhances solution speeds.

Large constellations of small satellites in low Earth orbit (LEO) are expected to play a significant role in Earth observation services in the future. One important application area is wide-area maritime surveillance, which can provide continuous and dynamic information on ships contributing to enhanced maritime situational awareness. Unlike the information processing of only one satellite, ship tracking using multiple LEO satellites is a new research topic that requires an effective fusion of information from different sources. The paper introduces a modified DeepSORT tracking methodology that incorporates both motion and appearance features for multiple ship tracking across multiple LEO satellites. The effectiveness of the proposed techniques is verified through the case study area imaging by five commercial LEO satellites, which shows the application potential of the LEO satellite constellation in persistent maritime surveillance.

Since polarimetric synthetic aperture radar (PolSAR) terrain segmentation is a dense prediction task, the disadvantage of inadequate labeled samples greatly limits its performance. In this article, we present a semi-supervised segmentation network called SemiPSCN to reduce the data reliance on label annotation, which integrates semi-supervised learning (SSL) paradigm and the characteristics of PolSAR data into a unified architecture. First, considering the unreliability of pseudolabels caused by noise interference in PolSAR data, a pseudolabel error localization (PEL) module is designed. By mapping the pixels that have mispredictions in pseudolabels, PEL can greatly enhance the confidence of pseudolabels. Then, SemiPSCN introduces a category representation constraint (CRC) module to explicitly boost the category consistency between labeled and unlabeled PolSAR data. Via explicit intracategory and intercategory constraints, CRC can guarantee the invariant representations on the same category region between labeled and unlabeled data. Furthermore, a region consistency constraint (RCC) module is designed to enhance the regional consistency in PolSAR data. RCC leverages the conception of graph to model the understanding of spatial relationships among terrain targets, thereby facilitating consistent spatial region expression in semi-supervised process. Finally, we build a challenging large-scale dataset called LSPolSAR-Seg and conduct abundant experiments on LSPolSAR-Seg. SemiPSCN exhibits superior performance when compared with other advanced approaches, especially improving mean intersection over union (mIoU) by 3.44%–12.77% under 20% split setting, which promotes the performance to a state-of-the-art level.

Issues such as insufficient feature representation and background confusion make detection tasks for small object in remote sensing arduous. Particularly when the algorithm will be deployed on board for real-time processing, which requires extensive optimization of accuracy and speed under limited computing resources. To tackle these problems, an efficient detector called FFCA-YOLO(Feature enhancement, Fusion and Context Aware YOLO) is proposed in this paper. FFCA-YOLO includes three innovative lightweight and plug-and-play modules: feature enhancement module(FEM), feature fusion module(FFM) and spatial context aware module(SCAM). These three modules improve the network capabilities of local area awareness, multi-scale feature fusion and global association cross channels and space, respectively, while trying to avoid increasing complexity as possible. Thus the weak feature representations of small objects are enhanced and the confusable backgrounds are suppressed. Two public remote sensing datasets(VEDAI and AI-TOD) for small object detection and one self-built dataset(USOD) are used to validate the effectiveness of FFCA-YOLO. The accuracy of FFCA-YOLO reaches 0.748, 0.617 and 0.909(in terms of mAP50) that exceeds several benchmark models and state-of-the-art methods. Meanwhile, the robustness of FFCA-YOLO is also validated under different simulated degradation conditions. Moreover, to further reduce computational resource consumption while ensuring efficiency, a lite version of FFCA-YOLO(L-FFCA-YOLO) is optimized by reconstructing the backbone and neck of FFCA-YOLO based on partial convolution. L-FFCA-YOLO has faster speed, smaller parameter scale, lower computing power requirement but little accuracy loss compared with FFCA-YOLO. The source code will be available at https://github.com/yemu1138178251/FFCA-YOLO.

Spatial target pose estimation with deep learning has garnered increasing attention in recent years. However, the existing methods in this field suffer from poor generalization. In this study, we propose a robust and reliable pose estimation method for spatial targets. The method aims to achieve keypoints filtrating. It involves a detection network tasked with identifying the target area, while the subsequent stage employs a classification network to regress keypoints from the detected target area. To improve the accuracy of pose estimation, we leverage spatial target geometric constraints to formulate 2-D–3-D keypoints equations for an initial pose. Then, we create a nonlinear optimization equation based on the confidence of 2-D keypoints and accomplish nonlinear refinement. We conduct extensive experiments on commonly used datasets and demonstrate the effectiveness of the proposed method. Furthermore, thanks to the effectiveness of keypoints filtrating and nonlinear refinement, the proposed method is robust with challenging scenarios and domain bias.

Gaofen-4 (GF-4) satellite is the first high-resolution geostationary orbit (GEO) optical satellite in the world, which has appealing potential in wide-area maritime surveillance and can provide dynamic information of ships as its high revisit time. However, its spatial resolution is not high enough to obtain attributes of ships, while Gaofen-1 (GF-1) satellite in low earth orbit (LEO) can extract rich features, as its high spatial resolution. Therefore, the integration of GF-1 and GF-4 satellites can provide a better maritime situational awareness. The aim of the paper is to provide a feasible architecture of data fusion of GF-1 and GF-4 imagery for maritime surveillance, and a novel ship association method using multi-level information in order to improve the association accuracy, and experimental results verify the effectiveness of our method.