Conductive layer and alignment layer are the most common and important film structures in liquid crystal devices. The two film layers are located between the substrate and the liquid crystal, and the film transmittance decreases due to the refractive index mismatch problem, which reduces the light energy utilization rate of the liquid crystal device. To address this problem, this paper proposes a multi-measurement anti-reflection film design method based on conductive layer and alignment layer, and uses the conductive layer and alignment layer as a link in the film system to achieve high transmittance. First, a physical model of a multilayer anti-reflection film based on conductive layer and alignment layer is constructed. Then, based on this model, the corresponding film structure is theoretically designed to achieve high transmittance of the film. Finally, the influence of the liquid crystal power-on process on the film transmittance is systematically studied. The results show that the transmittance of the designed anti-reflection film remains above 99% in the 500~800 nm band and above 96% in the 400~500 nm band; driving the liquid crystal will cause the film transmittance to decrease. Based on the mixed refractive index of the liquid crystal, the transmittance decrease can be reduced to less than 0.5%. This study will be conducive to the realization of liquid crystal devices with high energy utilization and high stability.

This study proposes a liquid crystal polymer-based equal-power splitter. Through multi-angle polarized patterned photoalignment, precise control of liquid crystal molecular orientation in a two-dimensional plane was achieved, enabling the fabrication of a 1×2 equal-power splitter based on a Y-shaped splitter model. By optimizing the bending radius of the waveguide, two types of splitters with spacing of 80 μm and 160 μm were developed. Furthermore, these splitters were cascaded to construct a 1×4 equal-power splitter with uniform spacing. Experimental results demonstrate that parallel light transmission and equal-power distribution within the liquid crystal waveguides are achieved, providing an effective solution for enhancing the functionality of liquid crystal waveguide devices. This research offers a new approach for advancing the integration and miniaturization of liquid crystal polymer waveguides, with potential applications in optical computing and optical communication.

Metasurface devices greatly enrich the interaction between light and matter, promoting the development of planar optics. In recent years, with the rapid development of terahertz (THz) technology, dynamically tunable terahertz devices have been widely studied. Liquid crystal (LC) is a unique soft matter that combines the fluidity of liquids and the order of crystals, making it indispensable in the field of displays and spatial light modulations. The integration of liquid crystals and metasurfaces enables tunable terahertz devices. In this review, we summarize three types of liquid-crystal based terahertz metasurfaces and their research progress, including liquid crystal planar terahertz devices, liquid crystal integrated metal metasurfaces, and liquid crystal integrated dielectric metasurfaces. Furthermore, techniques for fabricating liquid crystal microstructures as well as liquid-crystal integrated metal or dielectric metasurfaces are introduced. Due to the excellent responsiveness of liquid crystal metasurfaces to external stimuli including electric fields, magnetic fields, heat and light, the amplitude, frequency, phase, and polarization state of terahertz waves are dynamically manipulated. The tunable terahertz metasurface devices based on liquid crystals drastically promote the application of terahertz in imaging, communication, and sensing fields.

The electro-optical performance of the viewing angle of polymer dispersed liquid crystal (PDLC) films can exert an influence on people’s perception effect in practical applications. To optimize the viewing angle performance of PDLC films and reduce the difference in haze between the side view and the front view, thereby improving people’s actual perception, this paper conducts a study on the electro-optical characteristics of PDLC films from the perspective of polymer monomers and explores the factors affecting the viewing angle performance of PDLC. PDLC is fabricated using the polymerization-induced phase separation (PIPS) technique. The viewing angle haze values of PDLC films fabricated with bifunctional monomers, methyl-functional monomers and hydroxyl-functional monomers are tested using a haze meter, and the difference in haze between the frontal viewing angle and the lateral viewing angle of PDLC is calculated. The experimental results indicate that the PDLC fabricated with polymer monomers such as EOHDDA, IBOMA and 4-HBA, which feature a moderate alkyl chain length, a refractive index (no) matching that of the liquid crystal, and low steric hindrance, has the smallest viewing angle haze difference (ΔH), which is 2.66%. The electro-optical performance of this wide-viewing-angle PDLC film is the best, with the smallest viewing angle difference and the optimal visual effect,which can meet the needs of practical applications.

Mini LED backlight display products use optical diffusion plates and composite films to improve brightness and uniformity. The test results of this paper show that the brightness gain of composite films ranges from 1.5 to 2 times, with the order of brightness gain being POP>DPP>MOP>DOP. This is due to the correction effect of different composite films on the beam solid angle. At the same time, the composite film raised the backlight color coordinates, with the change in the diffusion plate+DPP color coordinates being Δx of 0.01 and Δy of 0.018 7. Analysis shows that the significant loss of short wavelength blue light radiation energy is caused by Mie scattering of diffusing particles. The blue light reflected back from the multi-layer optical microstructure of the composite film to the quantum film will generate multiple photoexcitation phenomena, further increasing the amplification of the xy color coordinate values of the quantum dot model by the composite film. The color coordinate changes generated by the optical composite film are beneficial for adjusting the color temperature of the display product to a lower color temperature from the factory to ensure display brightness.

To address the problem of insufficient accuracy in LCD Mura defect detection due to low contrast and diverse scale differences, from the perspective of improving the model’s performance in detecting small-scale defects and weak defects, an improved YOLOv8n-based LCD Mura defect detection model, YOLO-D3MNet, is proposed. Firstly, the backbone and neck networks of the model are reconstructed through the introduction of the ConvNeXtv2 module, which improves the weak feature extraction capability of the model under the background of complex texture. Secondly, for the problem of insufficient cross-channel communication of feature information in the detection head module, an efficient decoupling head combining the channel shuffle strategy and depth-separable convolution is proposed to promote the information flow between different feature channels and reduce the model computation power requirement. Finally, to address the problem that the intersection and concatenation ratio metric based on prediction box and truth box is sensitive to the positional bias of small-scale defects, the normalized Gaussian Wasserstein distance loss function is introduced to provide more positive sample candidate boxes, which improves the model’s detection performance of Mura defects. The precision, recall and mAP50 of the improved YOLO-D3MNet model are 92.9%, 88.8% and 94.8%, respectively. Compared to the base model YOLOv8n, the precision, recall and mAP50 of the YOLO-D3MNet model are improved by 3.4%, 2.7% and 3.6%, respectively, while the GFLOPs of the model are reduced by 24.7%. Compared with mainstream target detection models such as YOLOv5n, the experimental results show that the YOLO-D3MNet model proposed in this paper has better performance in LCD Mura defect detection.

In order to solve the problems of poor scalability, low generality and high computational cost of the traditional method using single object CNN model, and optimize the performance of multi-objective method. In this paper, a single-stage network architecture for multi-objective 6D attitude estimation is proposed, and a multi-branch feature extraction decoder is designed to capture and aggregate detailed features effectively. This paper proposes a feature optimization and screening module, which filters input features to extract multi-scale features. Combining the above two, a new feature pyramid structure is designed to improve the overall performance of the network and improve the pose estimation effect of occlusion. The experiments are carried out on synthetic data set LINEMOD and Occluded LINEMOD. The results show that the proposed method has achieved significant improvement in the processing of blocked object scenes. Compared with the most advanced methods such as PyraPose, SD-Pose and CASAPose, the proposed method has increased the ADD/S-Recall index by 43.1%, 16.1% and 12%, respectively. It performed better when the number of targets is small, increasing performance by 17% when the number of targets is 4. The ablation experiment further verifies the effectiveness of each module. By introducing multi-branch feature extraction decoder, feature optimization and screening module, and feature pyramid structure, the proposed single-stage multi-objective network architecture can process any number of targets by training only one network, and can perform 6D pose estimation better under the condition of synthetic data. Experimental results verify the effectiveness of the proposed method.

In order to address the problem of missed and false detections caused by different object scales, densities, unclear details and especially numerous small objects in drone aerial images, a novel object detection method based on Adaptive Slicing Aided Inference (ASAI) is proposed. Firstly, an aerial image is input into an object detection network for initial inference. With the initial inference results, a window scoring mechanism is designed to locate the ambiguous targets in the input image, and the effective image regions are selected automatically for slicing to adapt to objects with different scales and densities. Then, the sliced images are sent to the object detection network for secondary inference. Finally, the two inference results are processed by an improved non-maximum suppression (NMS) algorithm to obtain the final detection result. Experimental results on typical datasets of VisDrone2019 and AI-TOD indicate that the proposed method improves the mAP metrics of typical lightweight object detection models including YOLOv7-tiny, YOLOv8n, YOLOv8s and YOLOv9-C, effectively improving the performance of object detection for aerial images.

To address issues of insufficient illumination and complex degradation caused by underwater scattering in low-light underwater image enhancement tasks, we propose a Retinex-based variable attention framework. An illumination-guided variable attention module is designed, which uses semantic information from brightness feature maps to enhance dark regions and improve the adaptive enhancement capability of the model. We construct a spatial-frequency domain feature fusion module, which enhances the prominence of image texture information through multi-scale feature extraction and inter-level feature fusion. A medium transmission module is introduced to further address image artifacts caused by underwater scattering during enhancement. The experimental results show that compared with other cutting-edge methods, our method has significant improvements in four objective indicators, with PSNR reaching 22.781 8, SSIM reaching 0.882 1, UCIQE reaching 0.614 6, and UIQM reaching 3.365 9. The enhanced images exhibit excellent visual quality, validating the effectiveness of our algorithm in improving the clarity of low-light underwater images.

In response to the problems of weak applicability, low registration efficiency and poor robustness of existing point cloud registration algorithms when registering point clouds from different scenes, this paper proposes an efficient registration algorithm for models, indoor scene and outdoor scene point clouds. Firstly, voxel grid filtering is used to downsample the point cloud, and intrinsic shape signatures (ISS) is used to extract point cloud features. Then, fast point feature histograms (FPFH) are used to describe the feature points, and random sample consensus (RANSAC) algorithm is used for rough registration of point clouds. Finally, the voxelized generalized iterative closest point (VGICP) algorithm accelerated by a graphics processing unit (GPU) is used to achieve precise registration. Experimental results show that in the three-dimensional model, indoor and low overlap outdoor point clouds with noise, the proposed algorithm achieves high registration accuracy while only consuming 0.118 s, 0.306 s, and 0.648 s, respectively. Compared with existing registration algorithms, the registration efficiency is improved by 79.12%, 82.41%, and 88.28%, respectively. The proposed algorithm has high registration accuracy and efficiency in different application scenarios, and has stronger applicability and higher robustness.

Aiming at the problem of poor detection accuracy caused by pedestrian crowding and occlusion in natural complex scenes, a dense pedestrian detection algorithm based on YOLOv7 with optimized weights is proposed. First, to address the occluded pedestrian feature extraction problem, the weights of the backbone network are redistributed by the algorithms for typical geometric figures of rectangle and circle. Measuring principles and algorithms of typical plane cross-space efficient multi-scale attention module with cross-spatial learning (EMA), and the correlations between different channel features are learned cross-dimensionally, which can enhance the model’s attention to the visible area of the pedestrian target. Second, to address the problem of high complexity of the detection model, the efficient lightweight connection module (ELCM) is designed to improve the model representation ability and speed up the training speed. Finally, a focused bounding box loss function, Focal-SIoU loss, is constructed, which focuses on suppressing low-quality samples and adds angular loss to improve the detection accuracy of the model. Experimental results demonstrate that the proposed algorithm achieves mean average precisions of 83.7% and 82.6% on the Wider-Person and Crowd Human datasets, respectively, showing significant advantages in dense crowded pedestrian detection tasks.

In order to solve the problems of difficulties and low detection accuracy caused by the small sample size and variable target scale of wildlife datasets, a few-shot wildlife detection (MS-FSWD) algorithm based on multi-scale context extraction was proposed. Firstly, the multi-scale context extraction module was used to enhance the perception ability of the model for wildlife at different scales and improve the detection performance. Secondly, Res2Net was introduced as a strong classification network of the prototype calibration module to correct the class scores output by the classifier. Then, the shuffle attention mechanism was added to the RPN to enhance the feature map of the target region and weaken the background information.Finally, using the Balanced L1 Loss as the localization loss function improves the target positioning performance.Experimental results show that compared with the DeFRCN algorithm, MS-FSWD improves the novel class AP50 by 9.9% and 6.6% respectively in the 1-shot and 3-shot detection tasks on the few-shot wildlife dataset FSWA. On the public dataset PASCAL VOC, MS-FSWD is increased by up to 12.6%. Compared with the VFA algorithm, in the 10-shot task of the PASCAL VOC dataset Novel Set 3, the novel class AP50 is increased by 3.3%.

Due to limitations in hardware platform computing power and storage resources, achieving energy-efficient and efficient convolutional neural networks (CNNs) by using embedded systems remains a primary challenge for hardware designers. In this context, a complete design of a heterogeneous embedded system implemented by using a system-on-chip (SoC) with a field-programmable gate array (FPGA) is proposed. This design adopts a cascaded input multiplexing structure, enabling two independent multiply-accumulate operations in a single DSP, reducing external memory access, enhancing system efficiency, and lowering power consumption. Compared to other designs, the power efficiency is improved by over 38.7%. The design framework is successfully deployed in a large-scale CNN network on low-cost devices, significantly improving power efficiency of the network model. The power efficiency achieved on the ZYNQ XC7Z045 device can even reach 102 Gops/W. Furthermore, when inferring the VGG-16’s CONV layers by using this framework, a frame rate of up to 10.9 fps is achieved, which demonstrates the framework’?s effective acceleration of CNN inference in power-constrained environments.