最新刊期
卷 34 ， 期 8 ， 2026
- 摘要：Ground-based optical telescopes play an important role in space domain awareness. To address the degradation of detection capability caused by sky background radiation and atmospheric turbulence， this paper proposed a speckle filtering method based on atmospheric turbulence characteristics. First， the method performed cumulative summation on sequentially stored multi-frame images to obtain a high signal-to-noise ratio （SNR） observation image through long integration time. Then， target speckles could be extracted from the accumulated image， with an appropriate extraction threshold， combined with connected component detection and size-based filtering. Subsequently， a direction independence speckle filter with matching size was constructed based on the extracted speckle patterns from the accumulated image. Finally， the speckle filter was applied to the latest single-frame observation image for spatial-domain filtering， enabling effective target extraction. Experimental results show that for the observation of faint space targets under strong atmospheric turbulence and sky background radiation， the proposed speckle filtering method can improve the detection capability of ground-based telescopes by more than 0.93 magnitudes. The speckle filtering method is characterized by simple computation ， fast processing and strong robustness， enabling ground-based optical telescopes to achieve better performance in space object detection.关键词：ground-based optical telescope;daytime detection;atmospheric turbulence;speckle filtering;SNR140|87|0更新时间：2026-04-28
- 摘要：A multidimensional optical pod integrating snapshot multispectral and thermal infrared imaging was designed and implemented to meet the demand for airborne remote sensing with multimodal cooperative detection. The system adopted a pixel-level mosaic filter-based snapshot multispectral imaging scheme. Through a dual-camera cooperative configuration， 18 discrete visible-near-infrared spectral channels were constructed， covering two spectral ranges of 480~600 nm and 650~800 nm. A 3×3 pixel-level mosaic sampling structure was employed. Under a 2×2 pixel binning strategy， the effective single-channel spatial resolution reached 680×510， with an equivalent pixel size of 6.8 µm， enabling video-rate multispectral acquisition. A thermal infrared imaging payload was synchronously integrated to achieve simultaneous multi-source data acquisition. For heterogeneous multimodal data， a processing framework combining geometric registration， spectral feature augmentation， principal component analysis-based dimensionality reduction， and unsupervised K-means clustering was established. Thermal infrared information was further introduced into the principal component feature space to enhance land-cover separability. Experimental results show that the center wavelengths of the multispectral channels agree well with the design values， with bandwidths of approximately 2% of the center wavelengths. Radiometric calibration exhibits near-unity linear fitting coefficients， indicating stable and reliable system response. In representative applications， the fusion of thermal infrared information effectively improves water boundary integrity， vegetation detection statistics， and land-cover classification performance. The intersection-over-union （IoU） increases by 2.7% and 6.6%， and the Dice coefficient increases by 1.7% and 4.5% in two typical scenarios. These results demonstrate that the proposed multimodal optical pod has strong potential for natural resource surveys and ecological environment monitoring.关键词：airborne remote sensing;multimodal cooperative detection;optical pod;snapshot multispectral imaging;pixel-level mosaic;thermal imaging;land-cover classification159|84|0更新时间：2026-04-28
- 摘要：To achieve rapid， non-destructive and accurate detection of vulcanization accelerator content in rubber products， this study adopted terahertz time-domain spectroscopy technology， combined with data augmentation and chemometric methods， to conduct quantitative analysis of vulcanization accelerators in multi-component rubber mixtures. Aiming at the problems of serious spectral overlap of rubber mixtures， small sample size which was prone to model overfitting and poor generalization ability， a data augmentation strategy based on data fusion and Least Squares Gaussian Fitting （LSGF） was proposed， and a quantitative model of Genetic Algorithm-optimized Support Vector Regression （GA-SVR） was constructed. To reduce data dimensionality and improve modeling efficiency， the Variable Space Iterative Shrinkage Approach （VISSA） was used to extract features from the original and augmented spectra. The results show that data augmentation can significantly improve the predictive performance of the model. Among them， the LSGF method has the best effect； after VISSA feature extraction， the model accuracy is further improved， and the correlation coefficient Rp of the LSGF-augmented data in the prediction set reaches as high as 0.982 6， with the RMSEP as low as 0.002 3 . This method can provide technical reference for rubber formula optimization and green and sustainable development of the industry.关键词：terahertz spectroscopy;data augmentation;least squares Gaussian fitting;data fusion;quantitative analysis81|71|0更新时间：2026-04-28
Modern Applied Optics
- 摘要：To meet the demand for nanometer-level displacement and microradian-level angular high-precision measurement of optical components in fields such as lithography systems， precision assembly， and space optics， and to address the limitations of existing methods in achieving multi-degree-of-freedom （DOF） synchronous and integrated measurement， this paper proposed and experimentally validated a measurement system based on multi-channel differential laser interferometry. By designing a seven-channel interferometer configuration and establishing a unified mathematical model， simultaneous solving of six-DOF pose parameters—including translational displacements and rotational angles—of optical components was realized. Simulation results indicated that， under ideal conditions， the root-mean-square （RMS） errors of translational measurements in the X， Y， and Z directions were better than 3.384 nm， while the angular measurement errors did not exceed 4.616 μrad. Experimental verification demonstrated that， in a static environment， the RMS stability of displacement and angular measurements reached 7 nm and 16.4 μrad， respectively. Under step inputs of -300 μrad and -500 μrad， the linear correlation coefficients between the system outputs and autocollimator reference values were 0.984 and 0.937， respectively， with residual RMS errors maintained within 44 μrad. The proposed system featured a compact structure， strong resistance to environmental disturbances， and high linearity， making it suitable for real-time pose monitoring in high-precision applications.Future work will focus on multi-DOF coupling error calibration and adaptability under dynamic environments to further enhance the practicality and reliability of the system.关键词：precision alignment;high-precision measurement;six-degree-of-freedom pose measurement;multi-channel differential laser interferometry75|84|0更新时间：2026-04-28
- 摘要：Perovskite quantum dot films demonstrate significant application potential in advanced display and lighting technologies. However， the electrohydrodynamic printing process was susceptible to environmental interference， resulting in poor printing stability and non-uniform droplet diameters， which limited the scalable array fabrication of quantum dot micro-films. In this study， a regulation method of ambient relative humidity on the diameter and ejection stability of printed MAPbBr3 quantum dot micro-films was constructed， and the effects of ambient relative humidity and applied voltage on the printed film diameter as well as the jet rheological process were investigated. Experimental results indicated that stable printing of MAPbBr3 quantum dot micro-films could be achieved under conditions of ambient relative humidity of 50%-75% and applied voltage of 2.0 kV. As the ambient relative humidity increased from 55% to 70%， the average diameter of the printed MAPbBr3 quantum dot micro-films was reduced from 198 μm to 134 μm， with CV values below 0.1. When the ambient relative humidity was increased to 75% under an applied voltage of 2.3 kV， the average diameter of the printed MAPbBr3 quantum dot micro-films could be further reduced to 69 μm. The regulation mechanism on the printing quality of MAPbBr3 quantum dot micro-films is beneficial for enhancing the large-scale fabrication level of perovskite quantum dot film arrays.关键词：electrohydrodynamic printing;perovskite quantum dots;ambient humidity;droplet ejection;Thin film deposition28|20|0更新时间：2026-04-28
- 摘要：To address the limitations of continuous motion and complex drive structures in permanent magnet type diamagnetically levitated micro planar motors， this paper proposed a single-sided driven three-degree-of-freedom （3-DOF） micro-actuator with a PM array. This design achieved a unified approach to structural simplification and multi-DOF decoupled control. The system employed ultra-thin flexible printed circuit （FPC） serpentine coils to generate an adjustable magnetic field via four independent current channels. Based on magnetic field analysis and an equivalent magnetic dipole model， a mathematical model between current and electromagnetic force was established， forming a 3-DOF open-loop decoupling control method. Experimental results demonstrate stable 3-DOF motion covering a wide XY-plane area of 7 mm×7 mm， while maintaining good linearity throughout the range. The positioning accuracy achieves ±8.232 μm and repeatability is ±5.690 μm. By finely adjusting the driving current， the system achieves a Z-axis resolution of 0.95 μm and a displacement adjustment range of approximately 23 μm. Compared to conventional active magnetic levitation platforms， this structure reduces volume by more than an order of magnitude， significantly lowers current requirements and manufacturing costs， and demonstrates excellent potential for high-precision micro-positioning applications.关键词：diamagnetic levitation technology;micro planar motors;ultra-thin FPC;harmonic analysis;open-loop control30|21|0更新时间：2026-04-28
- 摘要：Industrial robots have become important carriers for automated assembly due to their high flexibility. However， in assembly tasks that have strict requirements for the distance between critical points on the dimensional chain （such as peg-in-hole assembly）， the existing absolute positioning accuracy of robots often struggles to meet the requirements. Currently， the main methods for improving industrial robot accuracy include kinematic calibration and spatial interpolation. However， kinematic calibration often pursues optimal global accuracy and fails to sufficiently consider the high-precision requirements of local discrete points on critical assembly paths； spatial interpolation is prone to introducing cumulative errors during the coordinate system transformation process， making it difficult to achieve the required assembly precision. To this end， this paper proposed a distance error prediction model and compensation method to improve the accuracy of critical points of robots in assembly tasks. First， the kinematic model of the robot was established， its position error model was derived， and an error model with distance error as a constraint was constructed to avoid error accumulation caused by coordinate system transformation. Second， by establishing the error mapping relationship between joint space and task space， the similarity characteristics of distance error in space were quantitatively revealed， and the variogram was used to quantitatively describe this similarity， providing a theoretical basis for interpolation point planning. Then， a distance error interpolation prediction model was constructed in the task space， and a transformation method from scalar error to vector error was proposed to realize the synchronous correction of error magnitude and direction. Finally， accuracy tests and assembly application experiments were conducted on a standard 6-DOF industrial robot. The experimental results demonstrate that after applying the proposed compensation method， the maximum and average positioning errors of the robot are reduced to 0.10 mm and 0.04 mm， respectively. Compared with kinematic calibration and the inverse distance weighting （IDW） method， the positioning errors are reduced by 66.77% and 49.51%， respectively. Furthermore， the assembly application experiments indicate that the maximum and average assembly positioning errors decrease from1.16 mm and 1.11mm （before compensation） to 0.15 mm and 0.05 mm， respectively. These results verify the effectiveness of the proposed method in assembly applications， providing a novel compensation strategy for the high-precision operation of industrial robots.关键词：industrial robots;distance error;spatial interpolation;error similarity;assembly29|18|0更新时间：2026-04-28
Micro\/Nano Technology and Fine Mechanics
- 摘要：In order to address the challenges of incomplete identification and inaccurate judgment in land cover monitoring， while balancing integrity and practicality， a novel change detection method based on co-refinement of object-level fusion and graph cut with KPCA-DSFA was proposed， which used two registered high-resolution remote sensing images. Firstly， relative radiometric correction and band stacking fusion were performed on the two-phase images. A simple non-iterative clustering algorithm was adopted for joint segmentation to generate homogeneous blocks that preserved the feature consistency of both image phases. Then the kernel PCA convolutional mapping network and deep slow feature analysis were coupled together for spatial-spectral feature extraction and deep semantic parsing respectively. Taking super pixels as the basic processing units， object-level high-dimensional spatial vectors were constructed via feature fusion to obtain change confidence information. Finally， an energy function model was established based on the Graph Cut， which leveraged the adjacency relationships and spatial differences of super pixel objects to achieve precise extraction of change regions through global optimization. Experimental results demonstrate that the proposed method achieves an overall accuracy of over 90% with excellent comprehensive performance. It can effectively suppress "salt-and-pepper" noise， significantly improve the recall rate of change regions， and exhibit favorable superiority and robustness.关键词：kernel convolution mapping;change detection;super pixels;deep feature analysis;graph cut model26|19|0更新时间：2026-04-28
- 摘要：The complex of traditional calibration operation and difficulty in processing calibration reference object may lead to low calibration accuracy. To address these issues， a hand-eye calibration method was proposed by using registration for constrained contour point sets and 2D laser scanners. It was formulated by a planar right-angled triangle as a calibration reference object. Firstly， the calibration idea of the registration method with constrained contour point sets was elucidated. The calibration process was then introduced based on the designed asymmetric right-angled triangle reference object. Secondly， a constrained point sets registration model was established with the contour information of the triangle. It utilized a coarse registration method by corner points to narrow the spatial search range， followed by a precise registration according to the constraint relationship between feature points and contour lines. Finally， an optimal solution to fit the contour feature was obtained by the covariance matrix adaptive evolution strategy. Besides， the profile loss error evaluation criteria were introduced to assess the performance of the calibration results. The first calibration experimental results show that the average profile loss error under 12 robot pose combinations is 0.084 2， the average fitting deviation of the reconstructed sphere diameter is 0.018 5 mm， and the maximum projection error of the sphere is 0.065 mm， which indicate fine calibration stability， efficiency， and accuracy. The second experiment is carried out with three different poses of the calibration object. The calculated parameters are used to fit a standard sphere， and both the diameter fitting deviations and the standard deviation of projection error are lower than 0.1 mm. The mean standard deviation of the fitted spherical center coordinates is 0.101 mm， lower than the 0.148 7 mm obtained by the one-step method. It indicates high robustness， efficiency， and versatility， which can meet industrial calibration accuracy requirements.关键词：hand-eye calibration;line contour features;point sets registration;3D reconstruction29|18|0更新时间：2026-04-28
- 摘要：Traditional point cloud registration algorithms are often observed to converge to local optima. This occurs when initial pose errors， noise， and repeated structures exist. To address this issue， a registration method was proposed to integrate deep feature consistency constraints with an attention mechanism. An attention-enhanced deep feature extraction network （AENet） was constructed. It was trained via self-supervised learning to generate point-wise descriptors invariant to rigid transformations. Using these descriptors， initial correspondences were established. A coarse transformation was then estimated， providing a reliable initialization for subsequent refinement. A deep feature consistency term was embedded into a multi-scale iterative closest point （ICP） optimization framework. It formed a joint objective that unified geometric alignment and deep feature matching for coarse-to-fine registration. By incorporating the feature similarity constraint into geometric optimization， a unified model was established. This model jointly enforced geometric proximity and deep feature consistency throughout the registration process. Refinement was performed progressively. It proceeded through coarse， intermediate， and fine scales. This improved convergence stability and reduced sensitivity to challenging conditions such as large pose variations and structural ambiguities. Extensive experiments are conducted on the ModelNet40 dataset. Results demonstrate significant improvements across multiple error metrics. Specifically， the root mean square error of rotation （RMSE R） is reduced by approximately 85.5%， 89.7%， 78.8%， 74.3% and 61.6% compared to FINet， OGMM， IDAM GNN， PREDATOR and RoCNet respectively. Similarly， the root mean square error of translation （RMSE t） is lowered by about 88.2%， 87.0%， 51.1%， 72.0% and 18.2%. These results indicate that the proposed framework effectively improves both accuracy and robustness. It provides a practical solution for high-precision point cloud alignment. The method performs well under complex environments and structural ambiguities.关键词：point cloud registration;deep feature;attention mechanism;iterative closest point （ICP）;multi-scale optimization;self-supervised learning32|17|0更新时间：2026-04-28
- 摘要：To address the color distortion， low contrast， and scale-dependent blurring caused by absorption and scattering in the underwater medium， a U-shaped underwater image enhancement network integrating dynamic perception and progressive refinement is proposed. Firstly， a dynamic optimization Transformer module is designed in the encoder， which dynamically perceives non-uniform degradation in underwater space by extracting anisotropic features， thereby enabling spatial contextual detail enhancement and global color restoration. Subsequently， a progressive refinement architecture is constructed in the decoder， comprising a dilated convolution aggregation module and a progressive semantic-guided upsampling module. Specifically， the dilated convolution aggregation module utilizes a staircase dilation strategy to capture multi-scale contextual features， and integrates a channel shuffle operation and a residual attention mechanism to optimize blurred region reconstruction， thereby improving the overall contrast and color fidelity of the images. The progressive semantic-guided upsampling module adopts a cascaded gradient preservation strategy to compensate for high-frequency detail loss， achieving a dynamic balance between deep semantic reconstruction and shallow detail optimization. Compared to nine state-of-the-art algorithms across five public datasets， the proposed method achieves the best PSNR， SSIM， and MSE metrics on the UIEB and LSUI datasets. Relative to the second-best approaches， it represents improvements of 1.47% in PSNR and 1.26% in SSIM， alongside an 8.33% reduction in MSE on the UIEB dataset. On the LSUI dataset， the PSNR and SSIM are improved by 14.51% and 6.92%， respectively， while the MSE is reduced by 45.00%. Furthermore， on the UCCS， U45， and C60 no-reference datasets， our method also yields the best results in terms of the UCIQE metric. These results validate the superiority of the proposed method in correcting color deviation， improving contrast， and recovering details， indicating that the generated images can serve as high-quality inputs for advanced visual tasks.关键词：underwater image enhancement;attention mechanism;dynamic perception;progressive refinement;color correction;feature fusion32|21|0更新时间：2026-04-28
Information Sciences
- Address：No.3888 Dong Nanhu Road, Changchun, Jilin, China Postal code：130033
- Tel：0431-86176855 Email：gxjmgc@ciomp.ac.cn
- Technical support is provided by Beijing Founder electronics co., LTD 吉ICP备11002662号-17
京公网安备11010802024621 - It is recommended to read the content of this site in Chrome&IE9+. Please switch to extreme mode in browser 360.
- Cookies We use cookies to help provide and enhance our service and tailor content. By continuing, you agree to the use of cookies.
0