Dynamic tasks or resources localization in distributed edge computing systems with inaccuracies in node positioning represent a critical research direction. Existing closed-form methods often introduce auxiliary variables to pseudo-linearize the performance models, followed by parameter estimation via weighted least squares (WLS). Although such a procedure is frequently refined in a subsequent step, the inclusion of extra variables tends to amplify estimation errors, causing the results to deviate from the Cramér–Rao lower bound (CRLB) under high-noise scenarios. To address this limitation, we propose a two-phase closed-form algorithm. In the first phase, extra variables are eliminated using an orthogonal projection matrix, and an initial solution is derived via least squares (LS). Since the omitted terms contain relevant information about task location and dynamics, this preliminary estimate remains suboptimal. A second phase is therefore designed to enhance its accuracy. The proposed estimator has potential for real-time implementation, because it avoids iterative convergence and provides deterministic computational steps. Both theoretical analysis and numerical simulations demonstrate that the proposed estimator achieves the CRLB under moderate Gaussian noise conditions. Simulation results further confirm the computational efficiency of the method and its superiority over existing closed-form alternatives.

On-orbit data from aerospace missions indicates that low frequency errors of opto-mechanical systems have become a bottleneck limiting the accuracy improvement of star trackers. To ensure the thermal stability of the optical axis and enhance its application performance in spacecraft, a multi-material thermoelastic compensation mechanism is designed by adopting a large diameter and long-focal-length optical structure, combined with matching the thermal expansion coefficients of optical and mechanical materials. Combined with finite element thermal-structural coupling analysis, a mapping model between temperature field and optical axis offset is established, and high-frequency thermal residual errors are compensated via an experimental post-processing filtering technique. Additionally, a separate installation scheme of lens hood is adopted to reduce the impact of system thermoelasticity. Taking the HST-A1 star tracker of Wuhan University’s satellite mission as the research object, an integrated design of CMOS image sensor and optical lens, as well as a separate installation scheme of lens hood, are proposed to conduct the design and verification of optical axis thermal stability. By establishing a thermo-mechanical-optical coupling simulation model, The optical axis offset characteristics under two configurations, namely integrated installation and separate installation, are compared and analyzed. The effectiveness of the scheme is evaluated by verifying mechanical performance, stray light suppression performance, and weight control. Results show that under the working conditions of lens hood temperature at 16.5 °C and star tracker body/bracket temperature at 15 °C: for the separate installation configuration, the maximum optical axis offset of Star Tracker 1 around the X-axis and Y-axis are 0.04′′ and 0.075′′, respectively; the maximum offsets of Star Tracker 2 around the X-axis and Y-axis are 0.036′′ and 0.02′′, respectively. This represents a simulated improvement of more than 80% compared with the integrated installation configuration (1.29′′ in X-axis and 0.36′′ in Y-axis).

Conventional optical cameras encounter numerous challenges in space target detection, including severe space radiation effects and motion blur. In contrast, the event camera, as a novel optical imaging sensor, exploits its asynchronous output to achieve high temporal resolution and a wide dynamic range, thereby exhibiting tremendous potential for space target detection. However, current research on event cameras remains limited and lacks a solid theoretical foundation. To address this gap, the present study introduces the calculation theory of blackbody radiation and incorporates key factors—such as target distance and reflective area—to establish a more precise sensitivity model for space target detection using event cameras. In addition, the critical parameters influencing sensitivity are analyzed in depth. Finally, ground based simulation experiments and nighttime star observation tests are conducted to verify both the feasibility of using event cameras for space target detection and the reliability of the developed computational model.

A star tracker is widely used as a high-precision attitude measurement device for spacecraft. It calculates attitude by extracting the magnitude and the position of presumed detected stars by a CCD/CMOS sensor and matching them with stars in the star catalog. The traditional star identification methods typically require the selection of specific anchor stars, which may cause insufficient identification accuracy as the number of stars used in the rough search is limited. In this paper, we propose a star identification method based on spatial projection, which starts with preprocessing. Then, a method for online expansion and reconstruction of the star catalog is proposed, which provides more stored star data. After the rough recognition and coordinate system transformation, the final identification is realized in the polar coordinate system. All the star points in the star image are identified, and the attitude information is obtained at the same time. The performance of the identification method is verified by real night sky experiments. Stray light experiments are also carried out to prove good noise immunity capabilities. Compared with the traditional subgraph isomorphism method, the proposed method makes it easier to adjust the number of recognizable stars in the field of view and better recognition of specific areas. The method is of great significance for future tasks such as attitude measurement, celestial navigation, remote sensing measurement, and space target observation and tracking.

Droop control is a typical method for achieving power-voltage control in DC systems. However, the movement of traction trains causes real-time changes in the resistance of the traction network, leading to variations in power sharing at traction substations (TSSs) determined by droop control, thereby impacting the voltage operation quality of the DC traction power supply system (TPSS). In this paper, an adaptive power-voltage hierarchical control strategy considering rational traction power sharing and voltage compensation based on the consensus algorithm is proposed to mitigate the impact of train movement. Firstly, to enhance the convergence speed of inter-station communication algorithms in the system layer control, an improved consensus algorithm with a convergence factor is proposed to calculate the average values of TSS state variables needed for the local layer control. Secondly, in the local layer control, the per-unit average power of TSSs is introduced to compensate for the impact of train movement on the droop coefficient. Furthermore, to reduce power loss in the traction network, a base quantity for virtual power is proposed for further compensation of the droop coefficient. Through dynamic compensation of the droop coefficient, a rational sharing of traction power according to rated capacity and power supply distance at TSSs is achieved. Moreover, the voltage average is utilized for secondary compensation of voltage deviations at TSSs. Finally, simulations based on a four-terminal DC TPSS have been established to validate the effectiveness of the proposed method in terms of power sharing and voltage control.

When hypersonic vehicles fly in near space, the flow field near the optical window leads to light displacement, jitter, blurring, and energy attenuation of the star sensor. This ultimately affects the imaging quality and navigation accuracy. In order to investigate the impact of aerodynamic optical effects on imaging, the fourth-order Runge-Kutta and the fourth-order Adams-Bashforth-Moulton (ABM) predictor-corrector methods are used for ray tracing on the density data. A comparative analysis of the imaging quality results from the two methods reveals their respective strengths and limitations. The influence of the optical system is included in the image quality calculations to make the results more representative of real data. The effects of altitude, velocity, and angle of attack on the imaging quality are explored when the optical window is located at the tail of the vehicle. The results show that altitude significantly affects imaging results, and higher altitudes reduce the impact of the flow field on imaging quality. When the optical window is located at the tail of the vehicle, the relationship between velocity and offset is no longer simply linear. This research provides theoretical support for analyzing the imaging quality and navigation accuracy of a star sensor when a vehicle is flying at hypersonic speeds in near space.

Detecting tiny infrared objects constitutes a significant difficulty within computer vision, requiring the detection and precise localization of minute objects within thermal imagery that typically span just a few pixels. The extremely limited target scale, coupled with the frequently cluttered backgrounds in infrared scenes, makes the task highly demanding. To address these difficulties, we introduce GRAM-NET, a deep learning model that improves detection performance through the integration of multiple specially designed modules. In particular, GRAM-NET integrates a Context-aware Multi-Branch Semantic Aggregation (C-MSA) block and a graph-based reasoning component. The C-MSA block adopts a parallel multi-branch extraction mechanism to obtain representations across diverse scales and semantic levels, while channel reweighting is applied to emphasize information conducive to small target discrimination, thereby boosting saliency and reducing both noise and misclassification. Meanwhile, the graph reasoning component leverages structured relational modeling to capture long-range dependencies, further combining multi-scale graph construction with bidirectional fusion for improved cross-scale awareness. Extensive experiments on NUDT-SIRST, NUAA-SIRST, and IRSTD-1K indicate that GRAMNET achieves superior detection performance compared with current leading approaches.

Star sensors, as the most precise attitude measurement devices currently available, play a crucial role in spacecraft attitude estimation. However, traditional frame-based cameras tend to suffer from target blur and loss under high-dynamic maneuvers, which severely limit the applicability of conventional star sensors in complex space environments. In contrast, event cameras—drawing inspiration from biological vision—can capture brightness changes at ultrahigh speeds and output a series of asynchronous events, thereby demonstrating enormous potential for space detection applications. Based on this, this paper proposes an event data extraction method for weak, high-dynamic space targets to enhance the performance of event cameras in detecting space targets under high-dynamic maneuvers. In the target denoising phase, we fully consider the characteristics of space targets’ motion trajectories and optimize a classical spatiotemporal correlation filter, thereby significantly improving the signal-to-noise ratio for weak targets. During the target extraction stage, we introduce the DBSCAN clustering algorithm to achieve the subpixel-level extraction of target centroids. Moreover, to address issues of target trajectory distortion and data discontinuity in certain ultrahigh-dynamic scenarios, we construct a camera motion model based on real-time motion data from an inertial measurement unit (IMU) and utilize it to effectively compensate for and correct the target’s trajectory. Finally, a ground-based simulation system is established to validate the applicability and superior performance of the proposed method in real-world scenarios.

Conventional vision-based sensors face limitations such as low update rates, restricted applicability, and insufficient robustness in dynamic environments with complex object motions. Single-pixel tracking systems offer high efficiency and minimal data redundancy by directly acquiring target positions without full-image reconstruction. This paper proposes a single-pixel detection system for adaptive multi-target tracking based on the geometric moment and the exponentially weighted moving average (EWMA). The proposed system leverages geometric moments for high-speed target localization, requiring merely 3N measurements to resolve centroids for N targets. Furthermore, the output values of the system are used to continuously update the weight parameters, enabling adaptation to varying motion patterns and ensuring consistent tracking stability. Experimental validation using a digital micromirror device (DMD) operating at 17.857 kHz demonstrates a theoretical tracking update rate of 1984 Hz for three objects. Quantitative evaluations under 1920 × 1080 pixel resolution reveal a normalized root mean square error (NRMSE) of 0.00785, confirming the method’s capability for robust multi-target tracking in practical applications.

Reconstructing 3D scenes in a fully differentiable and photorealistic manner remains a longstanding challenge in computer vision, especially for optical environments involving transparency, reflection, and non-Lambertian surfaces. In this work, we present Dust3R-3DGS, a unified pipeline that combines Dust3R,(1) a transformer-based dense image matcher, with 3D Gaussian Splatting (3DGS),(2) a real-time, point-based differentiable rendering framework. Unlike traditional pipelines relying on Structure-from-Motion (SfM)(3) for geometry initialization, our method directly predicts dense 3D points and camera poses from raw image pairs. These predicted points are treated as the centers of Gaussian primitives. To provide a superior starting point for optimization, we introduce a lightweight MLP head that performs a one-time prediction of other initial attributes, including scale, rotation, and opacity, based on deep features from DUSt3R. The integrated system enables photometric supervision directly through differentiable rendering, allowing geometry and appearance to be optimized jointly. Although the current implementation keeps the Dust3R backbone frozen, the architecture is designed to support future end-to-end training. The core of our method lies in the subsequent optimization stage. Driven by photometric supervision from differentiable rendering, our system jointly refines all attributes of the 3D Gaussians and the camera poses. This joint optimization is crucial for correcting inevitable inaccuracies in the initial pose estimates and for achieving high-fidelity reconstruction of complex geometry. Our experiments on challenging scenes containing transparent and specular objects demonstrate that this approach robustly reconstructs high-frequency details and maintains realistic visual fidelity where traditional methods fail. Dust3R-3DGS provides a scalable and robust alternative for 3D modeling of complex optical scenes, bridging the gap between learning-based vision and differentiable rendering.

Star sensors, as the most precise attitude measurement devices currently available, play a critical role in spacecraft attitude estimation. For hypersonic vehicles equipped with star sensors operating in near-space, aero-thermal radiation effects from the flow field can significantly degrade the quality of star images captured by the star sensor, potentially leading to failure in attitude determination. To accurately assess the quality of star images under aero-thermal radiation interference, this paper proposes a star image simulation method based on aero-thermal radiation effects to construct a dataset. Combined with the gray-level co-occurrence matrix (GLCM) algorithm, an aerodynamic disturbance index (ADI) is developed to quantify the degradation level of star image quality. Experimental results demonstrate that the proposed method effectively evaluates the interference level of aero-thermal radiation on star images, providing valuable references for subsequent image denoising, star extraction, and spacecraft attitude determination.

In the development of autonomous driving perception systems, LiDAR-Camera extrinsic calibration serves as the core component for multimodal data fusion, where its accuracy directly determines the reliability of environmental perception and localization. Traditional calibration methods, relying on single optimization strategies (such as pure geometric feature matching or fixed parameter search), face issues of calibration parameters being trapped in local optima and insufficient robustness in complex scenarios, leading to constrained calibration accuracy. To address these challenges, this paper proposes a joint calibration framework integrating a multi-stage hybrid optimization algorithm with feature extraction: 1) designing a four-layer feature extraction architecture through adaptive multi-level LiDAR feature extraction to fuse geometric and semantic features; 2) constructing a multistage joint optimization pipeline that employs an improved RANSAC-PnP algorithm to obtain initial extrinsic parameters by fusing multimodal features, followed by Levenberg-Marquardt nonlinear optimization combined with a simulated annealing-based random search strategy and success rate feedback mechanism to avoid local optima; 3) conducting bidirectional reprojection verification to filter optimal parameters. Experiments on the KITTI dataset demonstrate that this method achieves 97.1% point cloud effective matching accuracy (16.42% improvement over traditional methods), while reducing both reprojection errors and extrinsic rotation matrix errors, effectively enhancing calibration precision in complex scenarios.

In order to investigate the effect of aero-thermal radiation on star sensor detection capability in near-space hypersonic flight platforms, this article proposes a star sensor simulation imaging technique that couples aerodynamic flow field thermal radiation with target star signals. First, a regional conical imaging sampling model is constructed to preprocess the flow field data, significantly improving computational efficiency. Second, a radiation transfer equation for nonequilibrium flow fields is introduced to establish an energy transfer model based on the star sensor's optical link. Finally, the detection performance of the star sensor under different flight conditions is quantitatively analyzed, and some simulation results are compared with flight test data for validation. The results indicate that the simulation model deviates by less than 30%. The method proposed in this article provides important data and theoretical support for star sensor navigation accuracy analysis and extreme condition assessment.

Most existing deep learning-based super-resolution (SR) methods for remote sensing images rely on predefined degradation assumptions (e.g., bicubic downsampling). However, when real-world degradations deviate from these assumptions, their performance deteriorates significantly. Moreover, explicit degradation estimation approaches based on iterative schemes inevitably lead to accumulated estimation errors and time-consuming processes. In this paper, instead of explicitly estimating degradation types, we first innovatively introduce an MSCN_G coefficient to capture global prior information corresponding to different distortions. Subsequently, distortion-enhanced representations are implicitly estimated through contrastive learning and embedded into a super-resolution network equipped with multiple distortion decoders (D-Decoder). Furthermore, we propose a distortion-related channel segmentation (DCS) strategy that reduces the network’s parameters and computation (FLOPs). We refer to this Global Prior-guided Distortion-enhanced Representation Learning Network as GDRNet. Experiments on both synthetic and real-world remote sensing images demonstrate that our GDRNet outperforms state-of-the-art blind SR methods for remote sensing images in terms of overall performance. Under the experimental condition of anisotropic Gaussian blurring without added noise, with a kernel width of 1.2 and an upscaling factor of 4, the super-resolution reconstruction of remote sensing images on the NWPU-RESISC45 dataset achieves a PSNR of 28.98 dB and SSIM of 0.7656.

Real-time target recognition and multi-motion parameters acquisition are critical in many applications. Conventional imaging methods struggle to achieve simultaneous high-precision recognition and motion analysis due to excessive data requirements, low update rates, and dynamic deformation interference. This paper proposes an image-free framework utilizing single-pixel detection, where six coding masks simultaneously compute affine moment invariants resistant to affine distortions (translation, rotation, scaling, shearing) and extract multiple motion parameters (centroid, rotation angle, scaling factor). The simulation results demonstrate that the affine invariants enable stable recognition of different aircraft. We further designed dynamic experiments to detect objects undergoing various affine transformations, which maintains robustness with a deviation coefficient of 2.12% and achieves accurate recognition of diverse targets at one-sixth of the DMD's maximum flipping frequency, approximately 3 kHz. Furthermore, the proposed method realizes a centroid localization accuracy of 1.12 pixels, a rotation error below 1.3°, and a scaling factor accuracy of 0.04. To our knowledge, this work represents the first demonstration of direct recognition of moving objects with shear distortion using single-pixel detection without performing image reconstruction. The proposed framework offers a novel solution for synchronous target recognition and motion analysis of high-speed targets with great potential in applications such as remote sensing and security monitoring, real-time tracking of fast-moving objects, and all-optical computing.

With the sharp increase in the number of low-Earth orbit satellites, ground-based optical observation is faced with the challenges of detecting and stably tracking satellite targets with low SNR (signal-to-noise ratio) under complex starry sky backgrounds. Aiming at the missed detection problems of existing methods in scenarios of strong noise interference, stellar residues and target maneuvers, this study proposes a satellite target extraction and tracking framework integrating adaptive spatio-temporal filtering and dynamic matching: in the spatial domain, adaptive morphological filtering is designed to suppress non-uniform background noise, and inter-frame differencing and recursive filtering in the time domain are combined to eliminate static stars or stars with apparent motion; secondly, the Hungarian algorithm is used for cross-frame trajectory correlation to eliminate false alarms for the second time; finally, a Multiple model Kalman filter is used to establish a target motion model, which can automatically adapt to the target's motion mode, so as to realize the target's trajectory prediction and state update. Experiments based on measured data show that the method significantly improves the detection rate of faint satellites (SNR < 2), and can realize continuous tracking of variable-speed targets, providing an efficient and reliable technical approach for large-scale space target monitoring.

In the hypersonic near-space environment, complex aerodynamic conditions impose severe challenges on star sensor calibration and performance evaluation, markedly degrading the detection accuracy of star simulators. To address these challenges, we developed an integrated computational model that incorporates aero-thermal radiation effects within a ground-based simulation platform and embedded it in a high-precision dynamic star simulator. First, we established a mechanistic model of aerodynamic thermal radiation adapted to near-space flow fields based on the operating principles of the star sensor. Second, we engineered a star simulator with high dynamic range and adjustable angular velocity capable of reproducing near-space flight conditions in the laboratory. Third, we conducted quantitative simulations to assess how aerodynamic thermal radiation affects star-sensor detection capability and simulator reliability. Experimental results demonstrate that our approach effectively simulates and evaluates these effects, providing both a theoretical foundation and technical support for the design and optimization of star-sensor systems in hypersonic near-space environments.

Fast and accurate object classification has attracted significant attention in numerous applications. Traditional imaging-based methods often face challenges due to high computational loads and data redundancy, which limit real-time performance. Recent research has demonstrated the potential of non-imaging techniques that combine invariant features with single-pixel detection. This paper proposes an optimized modulation strategy based on Hu invariant, reducing the required number of illumination patterns from five to three while maintaining classification accuracy. The approach achieves a 67% increase in update rate, reaching a theoretical frequency of 7.4 kHz. We performed recognition on digit images from the MNIST dataset, and simulation results showed that the proposed method attained a recognition accuracy above 90%, comparable to traditional methods. The proposed method offers a practical solution with potential applications in optical target classification, dynamic light field analysis, and real-time industrial inspection.

Real-time localization and classification of fast-moving objects are crucial in various applications. Traditional imaging approaches face significant challenges, including large data requirements, limited update rates, motion blur, and restrictions in non-visible wavelengths. This paper proposes an image-free method based on complementary single-pixel detection and centralized geometric moments, which effectively integrates target localization and classification into a unified framework. By employing only four specific illumination patterns, the method can simultaneously determine the centroid position and shape of the target at an update rate of up to 5.55 kHz. Theoretical simulations verify the robustness of the proposed method under similarity transformations. Experimental results indicate that the proposed system achieves accurate real-time target localization and classification under diverse conditions, with an RMSE for centroid localization below 0.5 pixels and 93.3% classification accuracy for 30 different objects. The proposed method demonstrates strong adaptability to complicated environments. It holds significant potential for applications in target tracking, character recognition, industrial automation, and the development of optoelectronic neural networks for advanced optical computing tasks.