最新刊期
卷 41 ， 期 5 ， 2026
- 摘要：Flexibly tunable liquid crystal optical elements exhibit broad application in smart displays， soft robotics， and information encryption. Cholesteric liquid crystals（CLCs）， endowed with Bragg reflection characteristics and excellent external field responsiveness by their chiral helical structures， are becoming strong competitors in this field. However， conventional CLCs usually rely on single external field regulation with limited tunability， and their high fluidity poses challenges for effective encapsulation and stable application on flexible substrates. To address these issues， this paper designs a thermosensitive liquid crystal system incorporated with photoswitch， while introducing a paintable ink composed of ethyl cellulose and terpineol to optimize viscosity， thereby achieving synergistic regulation via dual external fields of temperature and light. Experimental results show that the thermosensitive system enables broadband regulation of the reflection wavelength from 420 nm to 1 000 nm within the temperature range of 37 ℃ to 29 ℃. Upon incorporation of photoswitch， dynamic tuning of the reflection wavelength from 450 nm to 700 nm is achieved by light irradiation at a specific temperature， where temperature and light synergistically determine the pitch variation. Systematic optimization of the mass ratio between CLC and the ink reveals that the 9∶1 ratio strikes the best balance between brushability and film-forming integrity. This study successfully demonstrates photothermal synergistic dynamic regulation of the Bragg reflection wavelength of CLCs on flexible substrates， enabling on-demand customization of specific reflection colors， thereby providing a novel photonic platform with multi-external-field responsiveness and facile flexible encapsulation for information anti-counterfeiting and smart display applications.关键词：cholesteric liquid crystals;photothermal synergistic regulation;paintable ink;structural color;flexible optical applications329|203|0更新时间：2026-09-12
- 摘要：As innovative photoelectric composites characterized by unique microphase-separated morphologies， polymer-dispersed liquid crystals（PDLCs） eliminate the need for polarizers and overcome the inherent reliance of traditional liquid crystal devices on rigid cell structures. To address the critical challenges of high driving voltages and substantial electro-optical losses in current PDLCs， this review systematically summarizes the recent advances and underlying regulatory mechanisms of PDLCs incorporating polyhedral oligomeric silsesquioxane （POSS）-based monomers （POSS-SH and KH570-POSS） and their nanocomposite doping systems. Particular attention is devoted to elucidating how the incorporation of POSS monomers and nanofillers modulates the phase separation process， polymer network morphology， liquid crystal droplet size/orientation， and interfacial anchoring. Furthermore， the intrinsic structure-property correlations between these microscopic structural evolutions and macroscopic electro-optical performances—specifically driving voltage， contrast ratio， and optical loss—are comprehensively established. The findings demonstrate that the introduction of the unique cage-like POSS structure effectively optimizes the phase-separated polymer network， thereby significantly reducing the driving voltage and enhancing the contrast ratio. By deepening the theoretical understanding of performance modulation from the perspectives of microscopic morphological evolution and interfacial interactions， this review provides critical insights for the future design of high-performance flexible smart windows and advanced optoelectronic devices.关键词：Polymer dispersion liquid crystal;polyhedral oligosasiloxane;driving voltage;contrast ratio148|190|0更新时间：2026-09-12
Material Physics
- 摘要：Liquid crystal（LC） display technology maintains a dominant position in the panel market due to its mature manufacturing processes and superior cost-effectiveness. However， its response speed is still a critical factor restricting high-speed dynamic displays and optical modulation. While overdriving（OD） techniques can effectively shorten response times， they are frequently accompanied by the optical bounce （OB） phenomenon in the response curve， which severely degrades the quality of high-speed and high-precision displays. Based on Frank’s elasticity theory and the Ericksen-Leslie hydrodynamic equations， this study thoroughly elucidates the formation mechanism of the OB phenomenon by analyzing the dynamic response process of the LC molecular director under overdrive conditions. It is revealed that the overdrive voltage induces a spatially non-uniform distribution of molecular rotational velocities within the LC layer； the rotational velocity difference between the molecules in the bulk layer and those near the substrates triggers the OB phenomenon. Based on this mechanism， we propose an optimization technique to suppress OB by precisely modulating the overdrive voltage amplitude. Experiments using a 10 μm anti-parallel aligned LC-BYE7 nematic LC device were conducted to verify the simulation results. The results demonstrate that optimizing the overdrive voltage effectively mitigates the internal rotational velocity difference， reducing the depth of the OB by up to 57.2% and decreasing the response time by 18 ms， thereby effectively improving display performance. The method proposed in this study not only validates the accuracy of the theoretical model but also provides a technically viable pathway with practical engineering value for the application of LC devices in fields such as high-quality dynamic displays and rapid wavefront modulation.关键词：overdrive technology;optical bounce;director;response curve168|184|0更新时间：2026-09-12
-
Concentric Fresnel liquid crystal lens with ring bandwidth controlled by independent voltage driving
摘要：In order to solve the problem that the bandwidth of the outer zone of traditional Fresnel liquid crystal lens decreases sharply with the increase of the number of zones， which leads to the decrease of optical quality and the limitation of effective aperture， this study proposes a concentric ring Fresnel liquid crystal lens structure which uses interdigital electrode structure to independently control the driving voltage of Fresnel zone and then control the bandwidth of Fresnel zone， in order to broaden the bandwidth of the outer zone. Firstly， the driving voltage is limited in the linear response region of liquid crystal， and the driving voltage of each ring zone satisfies the derivative continuity condition through theoretical calculation to form a parabolic voltage distribution. Subsequently， a liquid crystal lens with an aperture of 2 mm， a liquid crystal layer thickness of 30 μm and three Fresnel zones （bandwidths of 0.4 mm， 0.3 mm and 0.3 mm） was fabricated. Compared with the traditional concentric ring design with maximum power， the outer zone increased by 5 μm and 73 μm respectively. Then， the polarization interference optical path is used to collect interference fringes and extract phase information. Experiments show that the goodness of fit of lens phase distribution and parabola is more than 99%， the maximum phase variation of positive and negative lenses is 20.9π and 22π， which is highly consistent with the theoretical reference value of 21.8π， and the power can be continuously adjusted in the range of -9.7~10.1 D， and it is approximately linear with the driving voltage difference. In addition， the imaging performance of the designed Fresnel liquid crystal lens is tested and evaluated. Based on ISO12233 resolution plate， the MTF50 value is 0.151 cycles/pixel，which is higher than that of Fresnel liquid crystal lens （0.132 cycles/pixel） designed by traditional formula and driven by double bus voltage. This result proves that increasing the bandwidth of Fresnel zone on the outer layer of the lens can effectively improve its optical imaging quality. The experimental results of zoom imaging also show that the liquid crystal lens has outstanding zoom ability. This design effectively realizes the expansion of the outer ring bandwidth and the parabolic phase distribution with high quality.关键词：optical devices;liquid crystal Fresnel lens;electrode structure;phase distribution193|296|0更新时间：2026-09-12 - 摘要：In liquid crystal（LC） lenses with composite dielectric layers， electrode extension is used to enhance the central electric field. However， this extension often makes the peripheral refractive index of the lens deviate from the ideal parabolic model. Consequently， the effective aperture is constricted to improve focusing quality， resulting in a loss of optical throughput. To optimize the refractive index distribution across the entire aperture， a biased-electrode LC lens based on the generalized Snell’s law is proposed. An inner-ring and a central biased electrode are added between the peripheral electrode. The structure effectively modulates the gradient electric field in the LC layer. Simulation results demonstrate that at
Device Physics and Device Preparation
- 摘要：To address the issues of low automation， insufficient detection accuracy， and difficulty in deploying on mobile devices due to large parameter counts in current cone recycling devices， this paper proposes a lightweight recognition algorithm named RGLU-YOLOv11n based on an improved YOLOv11n. First， RepConv and TransNet’s CGLU （Convolutional Gated Linear Unit） are integrated to reconstruct the C3K2 module， reducing model complexity and improving detection speed. Second， Ghost convolution and adaptive downsampling （ADown） are introduced to design a lightweight Ghost-ADown module， which notably reduces computational redundancy during downsampling while preserving feature richness. Third， a Lightweight Shared Convolutional Detection（LSCD） head， is introduced to achieve model lightweighting while maintaining detection accuracy， making the model more suitable for deployment on resource-constrained edge devices. Finally， to address the issues of weak generalization and slow convergence of the CIoU loss function during detection， Focaler-IoU and MPDIoU are introduced to design a Focaler-MPDIoU loss function， replacing the original loss function to improve model accuracy and accelerate bounding box regression. Experimental results show that compared with the baseline YOLOv11n model， the proposed model achieves improvements of 1.1%， 1.4% and 1.6% in precision， recall and mAP@0.5， respectively， while reducing parameters and model size by 35.3% and 41.5%， respectively. These results demonstrate that the proposed model maintains high recognition accuracy while significantly reducing network computational load， meeting the deployment requirements for mobile and embedded devices.关键词：cone detection;deep learning;multi-module fusion;lightweight model;loss function86|131|0更新时间：2026-09-12
- 摘要：To tackle attention noise， insufficient fine-grained spatial modeling， and computational redundancy in current Vision Transformer（ViT）-based cross-modal person re-identification（ReID） methods， a multi-mechanism fusion-enhanced framework is proposed. Building upon the ViT backbone， the proposed method leverages a multi-dimensional feature fusion strategy to optimize feature representation. Firstly， a Difference Attention Module（DAM） is introduced to compute the discrepancy between parallel attention maps， effectively filtering out high-frequency background noise and significantly enhancing the model’s focus on pedestrian targets. Secondly， a hierarchical Relation-Aware Module（RAM） is designed to adaptively conduct progressive local-to-global spatial relationship modeling. This design effectively overcomes ViT’s inherent limitation of losing spatial structural information when facing occlusions and severe pose variations. Finally， a Cross-Strip Window Attention（CSWA） mechanism is integrated. By dividing the feature maps into horizontal and vertical strips for parallel self-attention computation， the method reduces computational complexity to linear time while efficiently capturing long-range global dependencies. The superiority of the proposed method is demonstrated through extensive experiments. On the MSMT17 dataset， an mAP of 68.1% and a Rank-1 accuracy of 86.3% are achieved. On the challenging Occluded-Duke dataset， mAP and Rank-1 are improved by 5.5% and 6.3%， respectively. Furthermore， the standard ViT baseline is outperformed on both Market-1501 and DukeMTMC-reID datasets. These results indicate that the proposed framework successfully overcomes feature divergence and spatial information loss， thereby boosting robustness and generalization under occlusions and complex backgrounds.关键词：person re-identification;cross-modal;ViT;DAM;CSWA;RAM86|115|0更新时间：2026-09-12
- 摘要：To address the problem of insufficient detection accuracy in power system inspection due to the small size， weak characteristics， and susceptibility to complex background interference of insulator defects， an improved YOLOv11n object detection model is proposed. The model integrates three key optimization strategies into its backbone and neck networks to comprehensively enhance defect detection performance. The C3k2_DFF module is introduced， which achieves efficient dynamic feature fusion through an adaptive channel attention mechanism and cross-scale feature dynamic weighted fusion strategy. The C3k2_LS module is designed to augment global context information perception via a large-kernel attention pathway. The ConverseUp upsampling operator is developed and incorporated to replace conventional upsampling methods； based on a learnable deconvolution mechanism， it actively reconstructs high-resolution detail information during feature map scaling transformations， thereby mitigating information loss. Experimental results on a real-world insulator defect dataset demonstrate that the improved model significantly outperforms the original YOLOv11n baseline in critical defect categories， with the broken class Average Precision（AP） improving by 3.2%， the flashover contamination class AP by 1.1%， the overall mean Average Precision（mAP） by 1.5%， and Precision by 5.7%， while Recall remains stable. The model also maintains high detection accuracy for “insulator” targets， validating that the proposed improvements enhance defect recognition capability without compromising fundamental detection performance. This enhancement strategy effectively strengthens the model’s ability to identify insulator defects， providing more reliable technical support for automated power line inspection.关键词：small object detection;YOLOv11;insulator defect detection;dynamic feature fusion;upsampling operator95|109|0更新时间：2026-09-12
- 摘要：To address the challenges of small-sized targets， complex background， dense distribution of Printed Circuit Board（PCB） surface defects， and the difficulty of existing detection algorithms in balancing accuracy and speed， this paper proposes a PCB defect detection algorithm based on OverLoCK-HAFB-APKConv with Improved Matching（OL-HAIM）. First， an Overview-first-Look-Closely-next（OverLoCK） backbone network is adopted， which leverages a Deep-level Decomposition Strategy（DDS） and Context-Mixing Dynamic Convolution（ContMix） to reduce computational complexity while enhancing feature representation for small targets. Second， a Hierarchical Attention Fusion Block（HAFB） is designed to replace the CNN-based Cross-scale Feature Fusion（CCFF） structure， exploiting a local-global attention mechanism to improve the detection capability for minute defects. Finally， an Adaptive Pointwise Kernel Convolution （APKConv） is introduced， which achieves adaptive deformation of the receptive field via learnable offsets to accurately capture defects of diverse morphologies. Experimental results demonstrate that OL-HAIM attains a mean Average Precision at IoU threshold of 0.5（mAP@0.5） of 97.6% on the PCB-DEFECT dataset， representing a 4.8 percentage-point improvement over the baseline， with only 13.8M parameters and Floating Point Operations Per Second（FLOPS） reduced to 23.6G， while achieving an inference speed of 55.7 frames/s（FPS） on the RK3588 embedded platform. Cross-domain experiments on the Northeastern University Surface Defect Detection Dataset（NEU-DET） and Microsoft Common Objects in Context（MS COCO） datasets further verify the generalization capability of the proposed algorithm. On field-collected samples from an actual PCB production line， the model achieves an mAP@0.5 of 95.1% and an inference speed of 55.4 FPS on the RK3588 embedded platform， concurrently meeting the requirements of high-accuracy detection and real-time performance in practical scenarios. The proposed method provides an effective solution that reconciles high accuracy with lightweight deployability for real-time PCB defect inspection in industrial settings.关键词：PCB defect detection;DEIM;lightweight object detection;edge deployment113|139|0更新时间：2026-09-12
- 摘要：In dense object detection scenarios， conventional defection methods often suffer from the loss of high-frequency edge information， blurred details during upsampling， and insufficient scale adaptability. To address these issues， this paper proposes a frequency-guided object detection algorithm， termed FW-YOLO. Specifically， a Frequency-Guided Omni-Dimensional Dynamic Convolution（FGODC） module is designed to decompose features by means of the Haar wavelet transform， while high-frequency responses are further exploited to guide convolution kernel selection， thereby enhancing the preservation of edge information during downsampling. In addition， a Frequency-Residual Guided Upsampling（FRUpsample） module is introduced to inject high-frequency features into the upsampling path， enabling the recovery of fine edge details. Furthermore， a Frequency-Guided Shared Head（FGSHHead） is constructed by combining cross-scale weight sharing with the FiLM mechanism， so as to improve the detection capability for objects at different scales. Experimental results on the CrowdHuman dataset show that the proposed method improves Precision， Recall， mAP50， and mAP50:95 by 1.1%， 3.8%， 1.8%， and 2.0%， respectively. These results verify the effectiveness of the frequency-guided strategy in dense object detection and provide a new perspective for dense object detection tasks.关键词：object detection;frequency guidance;wavelet transform;dynamic convolution;upsampling;dense pedestrian detection4468|167|0更新时间：2026-09-12
- 摘要：To address the long-standing challenge of balancing visual fusion quality with semantic task accuracy in infrared-visible image fusion， this paper proposes a multimodal semantic segmentation enhancement framework based on a semantic prior. While existing infrared-visible fusion methods can improve the visual quality of the fused image to some extent， they often weaken the semantic information representation within the image， thus affecting the performance of downstream semantic segmentation tasks. To solve this problem， this paper introduces a pre-trained Segment Anything Model（SAM） as a frozen semantic prior model to guide multimodal feature learning during the fusion process， maintaining the consistency of semantic information. In the proposed framework， RGB and infrared images are first partially input into the SAM encoder to obtain multi-level visual feature representations. Simultaneously， a specially designed fusion module extracts complementary features between different modalities. At each feature stage， the features extracted by SAM are further optimized through a semantic preservation branch， and a semantically aware cross-modal enhancement module is further introduced for feature strengthening. This module improves the model’s responsiveness to different target categories by fusing text label embedding information and effectively enhances the segmentation accuracy of target boundary regions. Experimental results show that the proposed method has achieved significant improvements in semantic information preservation， fusion accuracy， and visual reconstruction quality， verifying the effectiveness and stability of the framework in infrared-visible semantic segmentation tasks.关键词：multimodal semantic segmentation;infrared-visible fusion;pretrained SAM;feature modulation;gated fusion;semantic prior;cross-modal learning68|90|0更新时间：2026-09-12
- 摘要：Although Transformer-based trackers have demonstrated strong accuracy owing to their superior global modeling capability， they still suffer from large model size and high computational latency. To address these issues， this paper proposes TPTTrack， a lightweight Transformer-based single object tracking method built upon target prompt tokens. Specifically， a target prompt token is introduced and jointly fed with the target template and search-region features into a lightweight Transformer backbone， where target priors are exploited to guide cross-region feature interaction and enhance tracking robustness. In addition， a hierarchical attention decoupling mechanism is designed to perform feature extraction and fusion in shallow layers while reducing redundant self-attention computation in deep layers， thereby lowering the overall computational cost. A lightweight autoregressive sequence prediction module is further incorporated to enable dynamic target-state modeling and efficient state estimation. Comparative experiments on the GOT-10k and LaSOT benchmarks show that TPTTrack improves Average Overlap on GOT-10k by 0.9% and precision on LaSOT by 1.4% over the strongest real-time baseline. Meanwhile， the model contains only 5.5M parameters and 1.79G FLOPs， and achieves inference speeds of 111.6 FPS on GPU and 20.76 FPS on CPU， with the CPU speed being 10.8% higher than that of the strongest real-time baseline. Overall， compared with existing lightweight tracking algorithms， the proposed method achieves superior performance in terms of accuracy， speed， and deployment friendliness， providing an effective lightweight solution for single object tracking in resource-constrained scenarios.关键词：single object tracking;Transformer;lightweight model;target prompt token;hierarchical attention decoupling103|79|0更新时间：2026-09-12
Image Processing
- Address：No.3888 Dong Nanhu Road, Changchun, Jilin, China 130033 Postal code：130033
- Tel：0431-86176059 Email：yjxs@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