Unmanned aerial vehicles (UAVs) and birds are typical low-altitude small targets in optical remote sensing, often occupying only a few pixels and exhibiting highly similar appearances, which limits the effectiveness of appearance-based discrimination at long distances and low resolutions. To overcome this, we propose a non-appearance-based classification framework using kinematic and time–frequency features. At the trajectory level, kinematic features—including the coefficient of variation of velocity and acceleration, the Spatiotemporal Box-counting Fractal Dimension (SBFD), and the Local Higuchi Fractal Dimension (LHFD)—quantify multi-scale trajectory complexity. At the scale-variation level, time–frequency features, specifically the Time-Frequency Aware Singular Value Entropy (TF-SVE) derived from bounding-box area sequences, capture non-stationary oscillations from bird wing flapping, reflecting behavioral differences from rigid UAV motion. Experiments on a complex real-world dataset show that stacking these features achieves 99.47% classification accuracy, demonstrating a robust, resolution-invariant, and practically effective approach for non-appearance-based recognition of low-altitude targets.

Infrared small target detection remains challenging due to low signal-to-noise ratio and complex background interference. Existing methods mainly rely on single-frame enhancement or temporal fusion, while lacking explicit modeling of cross-domain statistical discrepancies in multi-frame sequences, which limits their robustness in complex scenarios. To address this issue, a multi-domain statistical consistency modeling framework is proposed to jointly characterize spatial structure, temporal correlation, and frequency stability. The key idea is to model the statistical consistency of background and the inconsistency of target responses across domains. Specifically, contrast-guided directional convolution (CGDC) suppresses directional background structures in the spatial domain, temporal low-rank decomposition (TLRD) models background consistency while enhancing cross-frame anomalous responses, and temporal frequency consistency suppression (TFCS) measures temporal fluctuations of high-frequency components to adaptively suppress unstable interference. These modules collaboratively improve the discriminability between target and background under complex conditions. Experimental results on DAUB and IRDST-15K datasets show that the proposed method achieves 95.90% and 78.40% mAP50, outperforming SSTNet and DTUM by 5.50% and 6.92%, respectively. Moreover, the proposed method effectively reduces false alarms while maintaining high recall, providing an accuracy-oriented trade-off between detection performance and computational cost.

The detection of streaking space objects in wide field space observation images is essential for space environment awareness, as these images serve as the core data for target cataloging and collision risk assessment. However, this task is fundamentally challenged by the elongated and arbitrarily oriented morphology of space streaks and their typically low signal-to noise ratio (SNR). These challenges often lead to fragmented detections, posing a significant obstacle for accurate cataloging and tracking. To address these issues, this work introduces two key innovations. First, we propose a local adaptive thresholding method for streak segmentation that estimates thresholds along multiple directions and employs the median absolute deviation (MAD) to eliminate outliers, thereby enhancing segmentation accuracy for streaks of arbitrary orientations in cluttered backgrounds. Second, we develop an enhanced Density-Based Spatial Clustering Algorithm (DBSCAN) with geometric constraints for streak labeling. This method bridges gaps in fragmented detections and reconstructs discontinuous structures, addressing the limitations of traditional eight-connected component labeling. By explicitly modeling the streak geometry, the proposed DBSCAN variant integrates adaptive geometric constraints into cluster labeling, with key enhancements—including recursive line fitting, adaptive neighborhood radius adjustment, and density-weighted clustering— to achieve robust target identification within complex point sets. The experimental evaluation demonstrates that the multi-directional thresholding and geometry-constrained clustering components of the proposed framework provide enhanced capability for detecting faint streaks and suppressing false alarms.

Subpixel localization of Gaussian spot-like targets (e.g., stars and space objects) in electro-optical star images is critical for celestial positioning and spacecraft attitude measurement. This article presents a comprehensive approach addressing the significant degradation of localization precision caused by outliers and non-Gaussian intensity distributions. The main novelties include: 1) a Gaussian point spread function (PSF) model with a bias correction term to resolve the Gaussian/non-Gaussian tension, effectively addressing systematic errors in small spot targets; 2) an iterative PSF estimation framework that progressively refines localization accuracy; and 3) a critical control point (CCP) strategy that provides robustness against outliers through deterministic iterative refinement and dynamic hard selection, distinguishing it from conventional robust fitting approaches like IRLS (which uses continuous weighting) or RANSAC (which relies on random sampling). This method assumes a symmetric Gaussian PSF, specifically tailored for scenarios where this assumption holds (e.g., stare-mode celestial positioning and spacecraft attitude measurement with calibrated optics). While assuming a symmetric Gaussian PSF model, our method introduces a correction term to compensate for the non-Gaussian grayscale distribution arising from the discrete sampling of small spot targets. Within each iteration, the framework selects appropriate pixel points as CCPs based on the current model estimation and uses these selected points to construct a new least-squares problem to update the target's PSF parameters. The CCP strategy effectively excludes strong outliers from participating in solving the least-squares problem, thereby enhancing both noise robustness and localization accuracy. The method's efficacy is validated through comprehensive simulations and experiments on real star image sequences. The method demonstrates effectiveness across various conditions, though limitations exist for severely undersampled targets, saturated pixels, anisotropic PSFs, and crowded star fields. Results indicate that the proposed method achieves competitive performance compared to existing approaches in terms of localization accuracy, noise robustness, and consistency across sequences.

Real-time dynamic star image registration is essential for star pattern recognition and target tracking. To ensure real-time registration within limited computing resources, a semianalytical Fourier-Mellin method is introduced for estimating registration parameters. The proposed method uses the position information and connected-domain areas of extracted star points to calculate the Fourier spectra, in contrast to the conventional pixel-level Fourier transform. Consequently, the method demonstrates exceptional computational efficiency and resilience to amplitude fluctuations in deep-space backgrounds. In addition, an innovative 1-D cross-correlation function is proposed for determining registration parameters, significantly reducing the number of search operations and computational complexity compared to traditional 2-D functions. Furthermore, a global coarse and local fine positioning strategy is implemented to enhance the framework's efficiency while maintaining registration accuracy. Numerical experiments validate that the proposed scheme constitutes a robust and real-time process for star image registration, making it a practical solution for on-orbit platforms with constrained computing and storage capabilities.

The robust detection of small targets is crucial in infrared (IR) search and tracking applications. Considering that many state-of-the-art (SOTA) methods are still unable to suppress various edges satisfactorily, especially under complex backgrounds, an effective infrared small target detection algorithm inspired by modified fast saliency and the weighted guided image filter (WGIF) is presented in this paper. Initially, the fast saliency map modulated by the steering kernel (SK) is calculated. Then, a set of edge-preserving smoothed images are produced by WGIF using different filter radii and regularization parameters. After that, utilizing the fuzzy sets technique, the background image is predicted reasonably according to the results of the saliency map and smoothed or non-smoothed images. Finally, the differential image is calculated by subtracting the predicted image from the original one, and IR small targets are detected through a simple thresholding. Experimental results on four sequences demonstrate that the proposed method can not only suppress background clutter effectively under strong edge interference but also detect targets accurately with a low false alarm rate.

Aircraft detection technology plays a vital role in civilian applications, with significant attention being devoted to research on related algorithms in recent years. However, most existing research predominantly focuses on aircraft detection from a single top–down viewpoint, which constrains the applicability of detection technology across diverse scenarios. To overcome this limitation, we propose RMVAD-YOLO, a multi-view aircraft detection model built upon YOLOv8. First, we propose a novel Robust Multi-Link Scale Interactive Feature Pyramid Network (RMSFPN), which robustly extracts features of the same aircraft category from multiple views while enhancing feature differentiation between different aircraft categories. Second, we propose the Shared Convolutional Dynamic Alignment Detection Head (SCDADH), which enhances task interaction and collaboration by sharing convolutions between the classification and localization branches while simultaneously reducing the number of parameters, enhancing the model’s ability to deal with multi-scale targets. Additionally, to further leverage background information and enhance the model’s adaptability to multi-scale target variations, we incorporate the LSK Module into the backbone network. Finally, we propose the WFMIoUv3 loss function, which strengthens the model’s focus on challenging samples and improves detection robustness. Experimental results on the newly released Multi-Perspective Aircraft Dataset (MAD) demonstrate that RMVAD-YOLO achieves an accuracy of 90.1%, a recall of 76%, 84.8% mAP@0.5, and 70.5% mAP@0.5:0.95, while reducing parameters and delivering an overall improvement in detection performance compared to the baseline YOLOv8n. RMVAD-YOLO also performed well on the VisDrone 2019 dataset, further demonstrating its reliable generalization capabilities.

Optical image-based space debris detection is crucial for preventing collisions between spacecraft in orbit. However, identifying debris amidst low signal-to-noise ratios with crowded stellar environment remains challenging. To enhance faint debris detection within complex backgrounds, this study proposes a multiscale approach that leverages the morphological feature differences between tracked space debris and nontracked stars for their identification. The proposed adaptive multiscale morphological transformation demonstrates robustness against adjacent star and noise interference, achieving superior local background estimation for debris segmentation. Furthermore, we construct a feature representation based on the differential response of pointlike and streaklike objects to morphological operations. This representation exhibits significant discriminative power for objects of differing morphology, facilitating their recognition. The proposed method offers a streamlined, one-step process for acquiring space debris, in contrast to the traditional methods of segmentation followed by classification. Numerical experiments confirm that the method surpasses traditional techniques, achieving higher detection rates and fewer false positives, particularly for debris within dense stellar backgrounds.

Infrared small target detection is significantly challenged by residual high-intensity background edges and a low signal-to-noise ratio. These issues hinder accurate target differentiation from the background and heighten the risk of false alarms. To address these challenges, we propose a method that employs multidirectional local gravitational force (LGF) contrast combined with level-line connectivity (LLC) contrast. The LGF model integrates information from each pixel within the local region and introduces a new sigmoid function to reduce noise, enabling fine-grained gradient detection. The magnitude and orientation in this gradient can then be used to differentiate the target from the background. Considering that the target exhibits different gradient features in different directions, we further propose a multidirectional LGF contrast. This contrast utilizes the distribution characteristics of LGF magnitude to enhance the target and effectively suppress strong edges. In addition, to fully utilize the orientation information in the LGF, we designed the LLC contrast based on the spatial consistency of the target, increasing the difference between the target and the background. Finally, we propose a regional fusion technique to weight the two contrasts, improving background suppression while preserving target intensity. Experimental results demonstrate the effectiveness of our method in detecting targets within high-intensity edge backgrounds, complex textures, and noisy environments. Compared to other state-of-the-art methods, our method significantly improves detection accuracy.

Infrared (IR) small target detection is gradually attracting a cornucopia of attention and developing rapidly in the application of IR search and track (IRST) systems. However, it is often hard to extract small targets with adequate feasibility and reliability, especially under a variety of different complex backgrounds. To improve the detection performance of small target embedded in sophisticated clutters, a novel hollow side window filter (HSWF) is designed to cope with the background estimation problem. Moreover, the saliency map constructed by heterogeneity filter is utilized as a critical weighting factor during the process of HSWF. Finally, with the help of commonly adopted patch contrast measure (PCM), the small target could be easily identified from the Hadamard product of the corresponding response maps. The effectiveness and outstanding performance of the proposed method are validated through comprehensive experiments and analysis.

Patch-based methods improve the performance of infrared small target detection, transforming the detection problem into a Low-Rank Sparse Decomposition (LRSD) problem. However, two challenges hinder the success of these methods: (1) The interference from strong edges of the background, and (2) the time-consuming nature of solving the model. To tackle these two challenges, we propose a novel infrared small-target detection method using a Background-Suppression Proximal Gradient (BSPG) and GPU parallelism. We first propose a new continuation strategy to suppress the strong edges. This strategy enables the model to simultaneously consider heterogeneous components while dealing with low-rank backgrounds. Then, the Approximate Partial Singular Value Decomposition (APSVD) is presented to accelerate solution of the LRSD problem and further improve the solution accuracy. Finally, we implement our method on GPU using multi-threaded parallelism, in order to further enhance the computational efficiency of the model. The experimental results demonstrate that our method out-performs existing advanced methods, in terms of detection accuracy and execution time.

In order to solve the problem that ground-based optical telescope cannot detect space targets in the daytime, an infrared detection method is proposed in this paper. Firstly, from the perspective of spectral characteristics of space targets and skylight, the short-wave infrared band (SWIR) of 0.9~1.7μm is optimized, which reduces the brightness of the skylight, and avoids the cooling of the optical system. Secondly, the signal-to-noise ratio (SNR) of the target is improved by the time-domain multi-frame accumulation algorithm and the goal of daytime 'extended range' detection of optical telescope is achieved. Experiments on a 0.3-meter telescope show that the space targets, which SNR less than 1 are detected by adopting this method in the daytime, when the zenith angle of sun is 40°, exposure time is 40ms, and the number of the cumulative frames is 360, the limit detection ability is 10.8mV. The results of the experiments well verify the theoretical analysis, which provides a reference for the system design of the next generation of ground-based optical telescope.

The blind pixels and the flash pixels of infrared detector, which are easily to be detected as the target, increase the false alarm rate. An algorithm about the detection of blind pixel and flash pixel is proposed in this paper. This method is based on the characteristics that the gray values of the blind pixels in the sequence images are basically unchanged, there is a jump in the gray values of the flash pixels, and there are obvious differences between blind pixels and neighboring pixels in spatial domain, the method of joint processing in space-time domain is used to detect blind pixels. Compared with the traditional algorithm, this method effectively avoids the problem of target signal suppression in blind flash pixel detection. The experiment of a 0.3m telescope shows that this algorithm can effectively eliminate blind flash pixels in infrared images. In the experiment of dim target detection, combined with multi-frame energy accumulation algorithm, this algorithm has the ability to detect targets with signal to noise ratio (SNR) ≤ 1 and the brightness of 10.8mVisdetected, which lays a technical foundation for the follow-up ground-based optical telescope to carry out full-time detection.

Infrared small target detection plays a significant role in precision guidance and infrared warning systems. Although existing methods based on infrared patch-image (IPI) model have achieved good detection results, most algorithms provide high detection accuracy but with low real-time performance, which limits their use in practical applications. In this paper, we presents a GPU based parallel implementation to focus on real-time performance of the small target detection by using IPI model and various optimization strategies. The parallel implementation is first analysed in details. Then, the speed of the method is tested on the embedded GPU Jetson AGX Xavier. Finally, the running time of the traditional algorithm on the CPU is compared. Experiments show a speedup of 20 × over CPU implementation for images with a resolution of 1024 × 1024 pixels, which has great potential for real-time applications. Our acceleration strategy is also useful for other infrared image-patch based small target detection algorithms.

State estimation for stochastic systems with unknown inputs has been a research hotspot in recent years. Many research results including the augmented state Kalman filter, the two-stage Kalman filter, the optimal two-stage Kalman filter and the robust two-stage Kalman filter (RTSKF) have been developed by various researchers. Considering that unknown inputs sometimes vary linearly in practical engineering, this paper addresses the problem of state estimation for linear systems with linearly-varying unknown inputs. A self-calibration Kalman filter with linearly-varying unknown input (SCKF-LVUI) is proposed where the unknown input is estimated by exploring the information from the state equation and state estimates at previous steps. The derivation of the SCKF-LVUI is given and the state estimate are calculated as well as the corresponding covariance matrix. Furthermore, a simulation example is conducted and demonstrates that the presented SCKF-LVUI has high estimation accuracy and can be conveniently applied in engineering applications.

Infrared small target detection technology has sufficient applications in many engineering fields, such as infrared early warning, infrared tracking, and infrared reconnaissance. Due to the tiny size of the infrared small target and the lack of shape and texture information, existing methods often leave residuals or miss the target. To address these issues, a novel method based on a non-overlapping patch (NOP) joint l0-l1 norm is proposed with the introduction of sparsity regularized principal component pursuit (SRPCP). The NOP model makes the patch lighter in the first place, reducing time consumption. The adoption of the l0 norm enhances the sparsity of the target, while the adoption of the l1 norm enhances the robustness of the algorithm under clutter. As a smart optimization method, SRPCP solves the NOP model fittingly and achieves stable separation of low-rank and sparse components, thereby improving detection capacity while suppressing the background efficiently. The proposed method ultimately yielded favorable detection results. Adequate experiment results demonstrate that the proposed method is competitive in terms of background suppression and true target detection with respect to state-of-the-art methods. In addition, our method also reduces the computational time.

Semantic image segmentation, as one of the most popular tasks in computer vision, has been widely used in autonomous driving, robotics and other fields. Currently, deep convolutional neural networks (DCNNs) are driving major advances in semantic segmentation due to their powerful feature representation. However, DCNNs extract high-level feature representations by strided convolution, which makes it impossible to segment foreground objects precisely, especially when locating object boundaries. This paper presents a novel semantic segmentation algorithm with DeepLab v3+ and super-pixel segmentation algorithm-quick shift. DeepLab v3+ is employed to generate a class-indexed score map for the input image. Quick shift is applied to segment the input image into superpixels. Outputs of them are then fed into a class voting module to refine the semantic segmentation results. Extensive experiments on proposed semantic image segmentation are performed over PASCAL VOC 2012 dataset, and results that the proposed method can provide a more efficient solution.

Semantic segmentation on high-resolution aerial images plays a significant role in many remote sensing applications. Although the Deep Convolutional Neural Network (DCNN) has shown great performance in this task, it still faces the following two challenges: intra-class heterogeneity and inter-class homogeneity. To overcome these two problems, a novel dual-path DCNN, which contains a spatial path and an edge path, is proposed for high-resolution aerial image segmentation. The spatial path, which combines the multi-level and global context features to encode the local and global information, is used to address the intra-class heterogeneity challenge. For inter-class homogeneity problem, a Holistically-nested Edge Detection (HED)-like edge path is employed to detect the semantic boundaries for the guidance of feature learning. Furthermore, we improve the computational efficiency of the network by employing the backbone of MobileNetV2. We enhance the performance of MobileNetV2 with two modifications: (1) replacing the standard convolution in the last four Bottleneck Residual Blocks (BRBs) with atrous convolution; and (2) removing the convolution stride of 2 in the first layer of BRBs 4 and 6. Experimental results on the ISPRS Vaihingen and Potsdam 2D labeling dataset show that the proposed DCNN achieved real-time inference speed on a single GPU card with better performance, compared with the state-of-the-art baselines.

Binocular vision calibration is of great importance in 3D machine vision measurement. With respect to binocular vision calibration, the nonlinear optimization technique is a crucial step to improve the accuracy. The existing optimization methods mostly aim at minimizing the sum of reprojection errors for two cameras based on respective 2D image pixels coordinate. However, the subsequent measurement process is conducted in 3D coordinate system which is not consistent with the optimization coordinate system. Moreover, the error criterion with respect to optimization and measurement is different. The equal pixel distance error in 2D image plane leads to diverse 3D metric distance error at different position before the camera. To address these issues, we propose a precise calibration method for binocular vision system which is devoted to minimizing the metric distance error between the reconstructed point through optimal triangulation and the ground truth in 3D measurement coordinate system. In addition, the inherent epipolar constraint and constant distance constraint are combined to enhance the optimization process. To evaluate the performance of the proposed method, both simulative and real experiments have been carried out and the results show that the proposed method is reliable and efficient to improve measurement accuracy compared with conventional method.

It is difficult to implement three-dimensional (3D) measurement in small field of view (FOV) or confined space with traditional sensors, for they cannot be put into or operated flexibly in such circumstances. To solve the problem, a sensor constructed by an electronic endoscope and a pair of mirrors is designed, combining the flexible characteristics of the endoscope transmission wire and the advantages of stereo technology. The calibration of the sensor and two corresponding points matching methods are described. For applications as diameter measurement of 3-D circle, an optimization method is used which directly obtains the diameter using the recovered 3-D points. The experiments show calibration and diameter measurement are of high accuracy, which provide the potential of expanding computer vision applications particularly in small FOV and confined environments.