Shuai Wang

The widespread deployment of unmanned aerial vehicles (UAVs) in civil and commercial airspace has raised significant safety concerns, driving the demand for reliable and real-time Anti-UAV visual detection systems. However, existing deep learning-based detectors face substantial challenges in complex low-altitude environments, including drastic scale variations, severe background clutter, and weak feature representation of small UAV targets. Moreover, handcrafted Transformer-based architectures often lack adaptability across diverse scenarios and struggle to balance detection accuracy with computational efficiency. To address these limitations, this paper proposes AutoUAVFormer, a super-resolution guided neural architecture search framework for Anti-UAV detection. In contrast to conventional manually designed approaches, AutoUAVFormer leverages joint optimization of a Transformer-based detection objective and a super-resolution reconstruction objective to automatically identify a task-specific optimal network architecture for detecting UAV targets. Specifically, a unified search space is formulated by jointly embedding Transformer hyperparameters and Feature Pyramid Network (FPN) structures, facilitating end-to-end co-optimization of multi-scale feature fusion and global context modeling. To efficiently locate architectures that balance accuracy and computational cost, a three-stage pipeline, combining supernetwork training with evolutionary search, is employed. Additionally, we design a super-resolution auxiliary branch that operates only during training to enhance the model’s ability to learn fine-grained textures and sharpen edge representations of small targets, without introducing any inference overhead. Extensive experiments on three challenging Anti-UAV detection benchmarks, namely DetFly, DUT Anti-UAV, and UAV Swarm, confirm the superiority of AutoUAVFormer over current state-of-the-art methods, with mAP@0.5 scores reaching 98.6%, 95.5%, and 89.9% on the respective datasets while sustaining real-time inference speed. These results demonstrate that AutoUAVFormer achieves strong generalization and maintains robust Anti-UAV detection performance under challenging low-altitude conditions.

With the rapid development of deep learning, significant progress has been made in remote sensing image target detection. However, methods based on deep learning are confronted with several challenges: (1) the inherent limitations of activation functions and downsampling operations in convolutional networks lead to frequency deviations and loss of local detail information, affecting fine-grained object recognition; (2) class imbalance and long-tail distributions further degrade the performance of minority categories; (3) large intra-class variations and small inter-class differences make it difficult for traditional deep learning methods to effectively extract fine-grained discriminative features. To address these issues, we propose a novel remote sensing aircraft recognition method. First, to mitigate the loss of local detail information, we introduce a learnable Gabor filter-based texture feature extractor, which enhances the discriminative feature representation of aircraft categories by capturing detailed texture information. Second, to tackle the long-tail distribution problem, we design a dynamic feature hallucination module that synthesizes diverse hallucinated samples, thereby improving the feature diversity of tail categories. Finally, to handle the challenge of large intra-class variations and small inter-class differences, we propose a contrastive learning module to enhance the spatial discriminative features of the targets. Extensive experiments on the large-scale fine-grained datasets FAIR1M and MAR20 demonstrate the effectiveness of our method, achieving detection accuracies of 53.56% and 89.72%, respectively, and surpassing state-of-the-art performance. The experimental results validate that our approach effectively addresses the key challenges in remote sensing aircraft recognition.

In recent decades, few-shot object detection in SAR imagery has gained prominence as a major research focus. The unique imaging mechanism of SAR causes the model to suffer from foreground–background imbalance and inaccurate extraction of class prototypes for novel class instances. Therefore, we propose an innovative few-shot object detection algorithm for SAR images via context-aware and robust Gaussian flow representation. First, we design the Context-Aware Enhancement module to address the foreground–context imbalance problem by refining representative support features into fine-grained prototypes, which are deeply fused with query features based on the prototype matching paradigm. Second, we devise the Manifold Class Distribution Estimation module to address the difficulty of class distribution estimation and the fluctuation of class centers of the sparse novel class. Furthermore, we formulate the Category-Balanced Difference Aggregation module to model the relationship between the base class and the novel class, addressing the sensitivity of the model to the variance of the novel class instances. Finally, we design the Cosine Decoupling Module so that the aggregated features are executed only on the classification branch without affecting the precise localization of the target. Experiments based on SAR-AIRcraft-1.0 and the self-constructed MSAR-AIR dataset indicate that the fine-grained detection and identification performance of the novel class of airplanes can reach 32.90% and 55.26%, respectively, in the 10-shot and 50-shot cases. In addition, our method enables cross-domain detection for different scenarios and sample types and exhibits excellent generalization performance in data-sparse scenarios.

Dynamic vibration absorbers array (DVAA) is a newly developed and promising technique for vibration attenuation of integrally bladed disk (blisk) by mounting underneath the disk. In this paper, the vibration attenuation characteristics and energy dissipation mechanism of DVAA for tuned and mistuned blisk are parametrically studied, where the viscous and frictional damping are both considered. The lumped parameter model of a blisk is employed for the convenience of parametrical study. Analytical power flow formulas within the blisk–DVAA system are derived to characterize the dynamic interaction between blisk and DVAA. Four typical modes of a blisk with different nodal diameters and deformation characteristics are selected to evaluate the performance of DVAA. Then, the effects of the mass ratio, the frequency ratio and the damping ratio of DVAA on the vibration of tuned and mistuned blisk are addressed. Numerical results show that a light DVAA can significantly mitigate the resonant amplitudes of the tuned and mistuned blisk. Such damper is effective for the modes with different modal characteristics and can provide robust vibration attenuation performance against random mistuning.

The global position precision of High-definition (HD) Map mainly relies on GNSS Real-Time Kinematic (RTK) positioning. RTK solution cannot be fixed in long urban canyon or long tree-lined roads for the non-line-of sight (NLOS) of signal and multi-path effect of signal reflection, consequently, the localization error in relative localization system such as IMU/wheel encoder, visual-inertial odometry (VIO), Lidar-inertial odometry (LIO) will drift dramatically. To improve the large drift error under long range GNSS denied environment, a multi-sensor-fusion localization algorithm assisted by pole-like objects registration is proposed in this paper. Firstly, an efficient relative pose observation model for two distant states are built by registration of pole-like object is proposed, in which an effective registration algorithm between pole-like objects detected by point-cloud and that detected by semantic segmentation on image plane algorithm is proposed. Secondly, a multi-sensor-fusion localization system which incorporate pole-like objects registration is proposed to get more precise global positioning result. The evaluation of the method proposed in this paper was carried out by the data acquired from a HD Map acquisition vehicle, the result shows that relative positioning precision improved from 0.5% distance to 0.18% distance error, the max absolute positioning error improved from 0.42 meter to 0.18 meter whenever the RTK is unavailable in a tracking distance of 500 meters.

Autonomous driving requires vehicles to achieve fine detection of objects in the surrounding environment, especially living pedestrians. Nevertheless, in real world road environments there are living pedestrians and roadside portrait billboards. Existing vision-based object detection technologies fail to ac-curately distinguish living pedestrians from human figures. As an important sensor of autonomous driving system, mmWave radar has extra help to detect living pedestrians. In this paper, we extract the radar cross section (RCS) of the object from the low-cost mmWave radar signal as a distinguishing feature between living pedestrian and portrait billboard. Based on this observation, we propose a feature fusion network of mmWave radar and computer vision based on attention mechanism, and detect living pedestrians from fusion features. We implement the design with commodity mmWave radar IWR6843ISK-ODS and RGB camera Logitech Pro C920. The evaluation results show that our method effectively detects living pedestrians with an mAP of 97.7% and outperforms existing studies.

When adaptive optics is applied to target imaging, laser atmospheric transmission, etc., variable extension target wavefront detection is a technical challenge. And its detection accuracy directly affects the correction effect of adaptive optics.In order to explore the wavefront detection accuracy problem of variable extension targets. In this paper, a simulation model for wavefront detection of variable extension targets with correlated Hartmann was developed. The model was based on the commonly used optical system parameters, and it was also established by using the theories of Fresnel diffraction, Newton's imaging equation, the working principle of Shack-Hartmann wavefront detector, and wavefront recovery.We analyzed the effects of target distance and attitude variations on wavefront detection accuracy. The relationship curves between RMS and PV of wavefront recovered residuals and target distance and different attitudes were obtained in the simulation. And we further carried out the analysis for the intrinsic physical reasons of forming this relationship. The results show that target distance and attitude changes affected the extension of targets within the Hartmann subaperture. When the wavefront was recovered using the correlation algorithm, we obtained that the smaller the extension, the higher the wavefront detection accuracy within the relevant Hartmann detection accuracy. Therefore, it could be summarized that the smaller the extension of the target, the more similar it was to the point target, then the detection error introduced by the extension became smaller.

Triangular frequency-modulated continuous-wave (FMCW) laser radars (ladars) are extremely sensitive to vibration errors. An FMCW ladar 3D imaging system may suffer from severe vibrations and can use only one-period echoes for the ranging of each observation spot; consequently, it can provide only few measurement results. These vibrations may cause large errors because conventional vibration compensation methods are ineffective when applied to fast disturbances with limited measurement results. To solve this problem, we analyze the influence of vibrations on FMCW ladar ranging and propose a vibration compensation method based on an instantaneous ranging model for one-period triangular FMCW ladar signals. We first use a synchrosqueezing wavelet transform to extract time-frequency curves of the up- and down-dechirp signals and then build an instantaneous ranging model that can characterize local vibration errors. Based on the instantaneous ranges, we remove the disturbance vibration errors by taking the mean values of the instantaneous ranges and obtain the target range by using the triangular relations of the up and down observations. Experiments based on synthetic and real data verify the effectiveness of the proposed method and its superiority over the three-point method and Doppler shift method in compensating for vibrations with different frequencies and noise levels.