最新刊期
      卷 32 ， 期 2 ， 2024

        Modern Applied Optics

      • “科技新闻播报：一项针对原子力显微镜（AFM）系统的创新研究取得突破。该研究通过精密控制探针与探针夹装配位置，实现了更换探针后与原光路位置的一致性，省去了繁琐的光路调整步骤。这一系统采用光束偏转法监测探针位置与偏转，并利用高精度位移与角度调节平台进行精确调整。实验验证显示，装配的探针平均位置精度接近1.1µm，更换一致性探针仅需8秒。这一成果不仅简化了AFM系统重新校准光路的操作，而且为工业计量型AFM的操作与测量性能提升奠定了坚实基础。这一创新研究为AFM领域的发展开辟了新的方向，有望推动相关技术的进步与应用拓展。”
        ”
        ZHANG Baoliang,LIANG Wenfeng,YANG Tie,YU Peng
        Vol. 32, Issue 2, Pages: 137-147(2024) DOI: 10.37188/OPE.20243202.0137
        摘要：To solve the problem of complex and time-consuming optical path adjustment after probe replacement in Atomic Force Microscope(AFM) systems， this paper presents the first proposal to achieve the consistency of the replaced probe with respect to the original optical path position of the AFM system by precisely controlling the probe and probe holder assembly position， thus eliminating the need to adjust the optical path after needle replacement in AFM systems. The optical path consistency component of the system used the beam deflection method to amplify and monitor the probe position and deflection， and used a high-precision displacement and angle adjustment platform to adjust the orientation of the probe relative to the probe clip. The probe consistency effect was verified by physical construction， and the impact of probe position deflection due to Ultraviolet （UV） glue curing process； the impact of detector noise on the imaging quality of the AFM system when the probe is deflected by different amounts was systematically analyzed. The experimental results show that the average position accuracy of the probes assembled by the system is close to 1.1 µm， and it takes only 8 seconds to change the consistent probes in the AFM system. The system achieves high precision and consistent probe assembly， which greatly simplifies the operation of the AFM system to recalibrate the optical path， and can effectively improve the operation and measurement performance of industrial metrology AFM when combined with the automatic needle changing device.  
        关键词：Atomic Force Microscopy（AFM);probe assembly;beam deflection method;micron-level displacement adjustment   
        287
        |
        59
        |
        0
        <HTML>
        <L-PDF><WORD><Meta-XML>
        <引用本文> <批量引用> 48419079 false
        发布时间：2024-02-01
      • “一项关于多传感器光电系统轴一致性检测的研究取得了重要进展。该研究针对传统检测系统工作波段范围较窄、系统灵活性较低的问题，创新性地结合了光路切换和光热转换的思想，设计了一套宽光谱多传感器轴一致性检测系统。该系统采用卡塞格林反射式光学系统，覆盖从可见光到长波红外范围，实现了宽光谱的检测能力。通过步进电机驱动导轨上方的反光镜位置移动，系统能够灵活切换光路，提高了系统的灵活性和适应性。同时，采用镀有硫化铜的锗玻璃作为光热转换靶材，将短波长的光斑转换为热斑，再用长波红外探测器实现对各波段激光光斑图像的采集，有效扩大了检测系统的光谱覆盖范围。研究团队对该检测系统进行了全面的性能评估。在0.4~14 μm波段光谱范围内，系统表现出良好的像质，由像差引起的弥散斑直径均在9 μm以下，能量集中度较高。此外，系统检测精度达到0.1 mrad，并通过导轨往返运动重复精度实验和系统测量准确度实验验证了系统的可靠性，结果表明检测系统满足仪表准确度1.5级的要求。该检测系统结构紧凑，适用波谱范围广，为多传感器光电设备的轴一致性检测提供了有效解决方案。这一研究成果在光电检测领域具有广泛的应用前景，为相关技术的发展和进步提供了新的动力。”
        ”
        ZOU Yun,ZHU Yundong,WANG Jinsong
        Vol. 32, Issue 2, Pages: 148-157(2024) DOI: 10.37188/OPE.20243202.0148
        摘要：Optical axis consistency is an important indicator for measuring the performance of multi-sensor optoelectronic systems. In order to address the issue of narrow working spectral range and limited system flexibility in the multi-sensor axis consistency detection system，this paper designed a board-spectrum multi-sensor axis consistency detection system. The system based on the concepts of optical path switching and opto-thermal conversion. The system used a Cassegrain reflective optical system as the receiving and transmitting system in the range from visible light to long-wave infrared. A stepper motor was used to move the reflective mirror above the guide rail， allowing for the switching of the system's optical path. The copper sulfide-coated germanium glass was used to convert the short-wavelength spot into a hot spot as the photothermal conversion target. The long-wave infrared detector was used to collect the laser spot images of various spectral bands. The system could detect the spectral range of 0.4 - 14 μm. The analysis of the optical system's image quality shows that the root mean square （RMS） diameter caused by aberrations is consistently below 9 μm across various wavelengths. The system also exhibits strong energy concentration. Detection accuracy of the system is analyzed， and the maximum measurement error is 0.1 mrad. Experiments on repetitive precision involving the guide rail's back-and-forth movement， and accuracy measurement experiments on the system， verify the reliability of the system. The results indicate that the detection system meets the requirement of instrument accuracy level 1.5. The detection system has a compact structure， a wide range of applicable spectral， and the ability to perform axis consistency testing for multi-sensor optoelectronic devices.  
        关键词：wide spectrum;axis consistency;optical switching;photothermal conversion;copper sulfide   
        175
        |
        51
        |
        0
        <HTML>
        <L-PDF><WORD><Meta-XML>
        <引用本文> <批量引用> 48419047 false
        发布时间：2024-02-01
      • “针对接触式测量方法装夹定位后只能测量一种或两种参数、检测效率低等问题，有专家提出了一种创新的舵类结构件几何量误差和装配误差视觉检测方法。该研究通过计算机视觉系统获取舵轴和舵面区域图像，并利用图像预处理技术去除图像畸变和噪声。在边缘检测方面，研究团队对比了多种算子，最终选用Scharr算子提取舵轴图像边缘，Canny算子提取摇臂图像边缘，确保了边缘提取的清晰度和连续性。此外，该研究还采用霍夫直线和霍夫圆检测方法提取舵面边缘线特征、舵轴母线特征、摇臂圆轮廓特征，构建了几何量误差检测目标函数，并运用自适应遗传算法计算最优解。值得一提的是，该研究还结合了相机标定的内外参矩阵，实现了舵芯对称度、舵轴垂直度、摇臂夹角的快速检测。经过多次重复测量实验，对称度检测精度达到0.055 mm，垂直度检测精度达到0.225 mm，装配角度检测精度达到0.772°，完成单次检测耗时7 s。这一方法不仅提高了几何量误差检测精度和效率，还有助于推动舵类结构件成型、制造、在机检测的自动化和智能化水平提升。该研究成果为舵类结构件的质量控制和生产效率提升提供了新的解决方案，具有重要的工程应用价值和行业推广前景。”
        ”
        YANG Zeqing,PING Enxu,CHEN Yingshu,HU Ning,ZHANG Yi,JIN Yi,LÜ Yali
        Vol. 32, Issue 2, Pages: 158-170(2024) DOI: 10.37188/OPE.20243202.0158
        摘要：In view of the problems that existing contact measurement methods can only measure one or two parameters after clamping and positioning， and low inspection efficiency， and combining the machine vision inspection technology has outstanding advantages such as no contact， no damage， high degree of automation and safety and reliability， a geometric errors vision inspection method of the air rudder was proposed. Firstly， the images of rudder shaft and rudder surface area were acquired by a computer vision system， and image distortion was removed using the camera parameters and distortion parameters obtained from the camera calibration. Image pre-processing techniques were used to remove noise and reduce the impact of non-target elements in the detection environment on the detected object. In order to extract the rudder-like structural member edges more efficiently， Sobel operator， Scharr operator， Laplace operator and Canny operator were used to detect the edges of the image to determine the contours.In order to more effectively extract the edge and determine the contours of the air rudder， the Sobel operator， Scharr operator， Laplace operator， and Canny operator were used to detect the edge detection of images. The experimental results show that the edges of the rudder shaft image were sharper and more complete after the Scharr operator， and the edges of the rocker image were sharper after the Canny operator. Therefore， the Scharr operator was used to extract the edges of the rudder image and the Canny operator to extract the edges of the rocker image. Combining the characteristics of the testing elements， Hoff straight line and Hoff circle detection methods are used to extract rudder surface edge line features， rudder shaft bus features， and rocker arm circle contour features. And the reference elements of the rudder core symmetry， rudder shaft perpendicularity， and rocker arm pinch angle are determined. Construct an objective function for geometric quantity error detection and compute the optimal solution using an adaptive genetic algorithm. The measured values of rudder core symmetry， rudder axis perpendicularity and rocker angle are obtained from the internal and external reference matrix obtained from the camera calibration. Finally， the vision inspection software for geometric and assembly errors of air rudder parts has been developed， and the vision inspection experiment platform has been constructed to realize the rapid detection of geometric and assembly angle errors of air rudder parts. After several repetitive measurement experiments， the symmetry inspection accuracy reaches 0.086 mm， the perpendicularity inspection accuracy reaches 0.233 mm and the assembly angle inspection accuracy reaches 0.373°， and the detection time is 7 s.The experimental results show that the method not only improves the accuracy and efficiency of geometric errors detection， but also helps to improve the automation and intelligence of forming-manufacturing-in-machine inspection of air rudder parts.  
        关键词：air rudder;geometric errors;Vision inspection;Hough transform   
        112
        |
        64
        |
        0
        <HTML>
        <L-PDF><WORD><Meta-XML>
        <引用本文> <批量引用> 48418962 false
        发布时间：2024-02-01

        Micro/Nano Technology and Fine Mechanics

      • “针对长波红外差分干涉仪在低温工况下出现的干涉条纹畸变问题，某研究团队进行了深入研究。该研究团队首先分析了干涉条纹畸变的影响因素，并结合光-机-热耦合分析方法对干涉仪系统低温工作状态进行仿真。随后，设计了一种针对光栅元件的低温微应力动态稳定支撑安装结构，并通过结构优化实现了显著的性能提升。经过全系统低温验证试验，该优化结构成功将干涉条纹畸变量控制在2个探测器像元以内，与仿真计算结果高度一致。这一研究不仅验证了优化分析方法的有效性，还为提升反射式光学系统结构低温稳定性，提高系统工作能力提供了有力支持。该研究成果在光学仪器设计与制造领域具有重要意义，为相关领域的研究与应用提供了新的思路和方法。”
        ”
        WU Yang,FENG Yutao,HAN Bin,WU Junqiang,SUN Jian
        Vol. 32, Issue 2, Pages: 171-183(2024) DOI: 10.37188/OPE.20243202.0171
        摘要：The Long-wave Infrared Spatial Heterodyne Interferometer may have interference fringe distortion due to non-uniform stress acting on the optical components under cryogenic conditions， which will cause performance degradation of the interferometer system. To solve the problem of interference fringe distortion under cryogenic conditions， this paper analyzed the factors affecting interference fringe distortion based on the initial optical mechanical system of Long-wave infrared spatial heterodyne interferometer， and combined the optical-mechanical-thermal coupling analysis method to simulate the cryogenic state of the interferometer system. Then， a cryogenic micro-stress dynamic stable installation structure was designed for grating， which is the key component affecting fringe distortion. After the optimization of structure， the Root-Mean-Square（RMS） and Peak-to-Valley（PV） values of grating’s surface shape are 3.89×10-2 nm and 2.21×10-1 nm， respectively， which are five orders of magnitude lower than the initial structure analysis results. The simulated interference fringe distortion is less than 1 detector pixel. The cryogenic verification test of whole system shows that the optimized structure can effectively reduce the distortion of interference fringe， and the distortion is less than 2 detector pixels. The experimental results are highly consistent with the simulation results， which verifies the effectiveness of the optimization analysis method. The optimization analysis method has great significance and value for improving the structural stability and operating performance of the cryogenic reflective optical system.  
        关键词：micro-stress clamping;interference fringe distortion;cryogenic opto-mechanical structure optimization;optical-mechanical-thermal coupling analysis;surface shape fitting   
        139
        |
        60
        |
        0
        <HTML>
        <L-PDF><WORD><Meta-XML>
        <引用本文> <批量引用> 48418884 false
        发布时间：2024-02-01
      • “针对旋变测角误差的难题，科研团队提出了一种基于特征频率参考的二次谐波误差自校正方法。该方法深入分析了旋变测角误差的机理，创新性地实现了幅值调整和相位差调整，有效校正了旋变输出信号中的二次谐波误差。实验结果显示，该方法可将旋变二次谐波测角误差幅值降低78.5%，伺服系统速率波动量降低40.5%，显著提升了旋变的测角精度和伺服系统的转速控制稳定度。这一研究成果不仅为伺服系统的优化升级提供了有力支持，也为旋变测角技术的发展奠定了坚实基础，有望在工业控制、航空航天等领域发挥重要作用。”
        ”
        WANG Yingguang,ZHANG Jiyang,ZHANG Qiang,LU Ming,TIAN Limei
        Vol. 32, Issue 2, Pages: 184-192(2024) DOI: 10.37188/OPE.20243202.0184
        摘要：The resolver amplitude error and quadrature error are expressed in the angular velocity spectrum as the second harmonic of the rotational frequency， which is the main source of the resolver angle measurement error and affects the angular velocity control accuracy and stability of the servo system. In this paper， a self-correction method for the second harmonic error of resolver based on characteristic frequency reference was proposed. Firstly， the mechanism of the resolver error was analyzed， and the mutual irrelevance of the amplitude error and quadrature error was obtained， and it was proved that the second harmonic error correction could be realized by adjusting the amplitude and phase difference of the resolver output signal. Then， the amplitude corrector based on proportional amplification and the phase angle corrector based on cross-adjustment were designed between the resolver and the Resolver-to-Digital Converter （RDC）. Finally， according to the constant characteristic frequency of the error signal in the linear control system， the servo system was controlled at a constant speed， and the amplitude of the second harmonic frequency in the angular velocity spectrum was used as the reference to adjust the amplitude corrector and phase corrector respectively， to correct second harmonic errors. The experimental results show that the method can reduce the second harmonic angle measurement error of the resolver by 78.5%， and the speed fluctuation of the servo system can be reduced by 40.5%.The method realizes the self-correction of the second harmonic error of the resolver， and can greatly improve the measurement accuracy of the resolver and the rotation stability of the servo system.  
        关键词：angel position sensor;resolver;angle measurement error;self-correction   
        92
        |
        53
        |
        0
        <HTML>
        <L-PDF><WORD><Meta-XML>
        <引用本文> <批量引用> 48418882 false
        发布时间：2024-02-01
      • “机械臂动力学模型辨识领域取得了重要突破。针对超冗余机械臂动力学模型的精确辨识问题，研究团队提出了一种创新的半参数动力学模型辨识方法。该方法结合迭代优化和神经网络补偿，有效提高了参数辨识的准确性和效率。通过遗传算法优化回归矩阵条件数生成激励轨迹，建立了关节非线性摩擦模型，进一步提升了模型的物理可行性。实验结果表明，与传统算法相比，该方法显著降低了关节辨识力矩残差的均方根值，验证了其有效性和优越性。这一研究不仅为超冗余机械臂的精确控制提供了有力支持，也为机械臂动力学模型辨识领域的研究开辟了新的方向。”
        ”
        ZHOU Yufei,LI Zhongcan,LI Yi,CUI Jingkai,HE Shunfeng,SHENG Zhanyi,ZHU Mingchao
        Vol. 32, Issue 2, Pages: 193-207(2024) DOI: 10.37188/OPE.20243202.0193
        摘要：In order to achieve accurate dynamic model identification of the hyper-redundant manipulator， a semiparametric dynamic model identification method based on iterative optimization and neural network compensation was proposed. First， the dynamic model of the hyper-redundant manipulator and the base parameter set were introduced， joint nonlinear friction model was established， and the excitation trajectory was generated using genetic algorithm to optimize the condition number of the regression matrix. Second， the physical feasibility constraint of the manipulator dynamic model was established， and a two loops identification network was designed to identify the inertial parameters and joint friction model of the hyper-redundant manipulator based on the iterative optimization method. Finally， the BP neural networks were trained to obtain the semiparametric dynamic model of the hyper-redundant manipulator by using data set. A series of identification algorithms were compared and analyzed. The experimental results show that， compared with the traditional least squares algorithm and weighted least squares algorithm， the identification algorithm proposed in this paper can improve the sum of identify torque residual root mean square （RMS） of joints by 32.81% and 23.76%， respectively. The sum of torque residuals of the semi-parametric dynamic model is 23.56% higher than that of the full-parametric dynamic model. The identification results verify the effectiveness of the proposed identification method.  
        关键词：redundant manipulator;dynamic model identification;iterative optimization;semiparametric dynamic model   
        111
        |
        57
        |
        0
        <HTML>
        <L-PDF><WORD><Meta-XML>
        <引用本文> <批量引用> 48418880 false
        发布时间：2024-02-01

        Information Sciences

      • “科技新闻播报：在点衍射干涉测量领域，一项创新的相位解包方法引起了广泛关注。该方法基于空洞空间卷积，将自编码器结构与空洞空间卷积相结合，实现了对包裹相位图像的高精度、高效以及抗干扰检测。研究团队制作了大量多样化的数据集，用于训练和优化这一算法，使其能够准确识别包裹相位所在阶次，并快速处理得到高精度的解包结果。与传统的解包算法相比，该方法在时间和精度上都展现出了显著的优势。处理一幅图像的平均时间仅需0.035秒，远低于传统方法的1秒以上。同时，实验结果表明，该方法的解包结果与专业干涉图像处理软件的枝切法处理结果相近，均方根误差值为0.022 2 rad，面形拟合结果的峰谷差值仅为0.012 1λ，均方根差值也仅为0.004 2λ。这一突破性的研究为点衍射干涉图像处理的高精度相位解包提供了新的可行方案，有望在相关领域发挥重要作用。”
        ”
        WANG Tongmeng,GAO Fen,LI Bing
        Vol. 32, Issue 2, Pages: 208-220(2024) DOI: 10.37188/OPE.20243202.0208
        摘要：In order to meet the high-precision， high-efficiency， and anti-interference detection requirements of point diffraction interference measurement for phase unwrapping algorithms. Atrous Spatial Convolutional Networks -based phase unwrapping method for phase-shifted point diffraction interference images was proposed. By combining the autoencoder structure and the adaptive spatial convolution， higher phase unwrapping accuracy was achieved， and the degradation of the network model was effectively prevented， realizing controllable multi-scale feature extraction of the wrapped phase image. A large and diverse dataset of point diffraction phase data was used for training and optimization， which accurately identifies the order of the wrapped phase and quickly processed the wrapped image to obtain high-precision unwrapping results. The proposed method was applied to actual point diffraction interference images and compared with results from ESDI professional interference image processing software and other unwrapping algorithms. The results show that the unwrapping results have an RMSE value of 0.022 2 rad compared to the software processing results， with a surface fitting result PV difference of only 0.012 1λ and an RMS difference of only 0.004 2λ. In terms of time efficiency， it takes only 0.035 s on average to complete the processing of an image， while the traditional methods are all greater than 1 s. Compared to other methods， the proposed method exhibits fast and high-precision characteristics in unwrapping wrapped phase， providing a new feasible solution for high-precision phase unwrapping in point diffraction interference image processing.  
        关键词：interferometry;surface measurement;interference fringe;neural networks;phase unwrapping   
        129
        |
        48
        |
        0
        <HTML>
        <L-PDF><WORD><Meta-XML>
        <引用本文> <批量引用> 48418878 false
        发布时间：2024-02-01
      • “在医学图像融合领域取得了一项重要研究成果。针对现有基于生成对抗网络（GAN）的医学图像融合方法存在的训练不稳定、提取图像语义信息能力不足等问题，研究团队提出了一种双耦合交互式融合GAN（DCIF-GAN）。这一创新方案在融合肺部肿瘤PET/CT医学图像时表现出色，有效提升了图像质量。DCIF-GAN通过设计双生成器双鉴别器结构，实现生成器与鉴别器之间的耦合，并通过全局自注意力机制强化交互式融合。同时，引入耦合CNN-Transformer的特征提取和重构模块，显著提高了对同一模态图像内部局部和全局特征信息的提取能力。此外，跨模态交互式融合模块的应用，进一步整合了不同模态间的全局交互信息。实验验证表明，DCIF-GAN在平均梯度、空间频率、结构相似度、标准差、峰值信噪比和信息熵等指标上，相比其他四种方法中最优方法，分别提高了1.38%、0.39%、29.05%、30.23%、0.18%和4.63%。这一成果不仅突出了病变区域信息，还使融合图像结构更加清晰，纹理细节更丰富。这一研究为医学图像融合领域提供了新的解决方案，有望为计算机辅助诊断技术的发展开辟新方向。”
        ”
        ZHOU Tao,CHENG Qianru,ZHANG Xiangxiang,LI Qi,LU Huiling
        Vol. 32, Issue 2, Pages: 221-236(2024) DOI: 10.37188/OPE.20243202.0221
        摘要：Medical image fusion based on Generative Adversarial Network （GAN） is one of the research hotspots in the field of computer-aided diagnosis. However， the problems of GAN-based image fusion methods such as unstable training， insufficient ability to extract local and global contextual semantic information of the images， and insufficient interactive fusion. To solve these problems， this paper proposed a dual-coupled interactive fusion GAN （DCIF-GAN）. Firstly， a dual generator and dual discriminator GAN was designed， the coupling between generators and the coupling between discriminators was realized through the weight sharing mechanism， and the interactive fusion was realized through the global self-attention mechanism； secondly， coupled CNN-Transformer feature extraction module and feature reconstruction module were designed， which improved the ability to extract local and global feature information inside the same modal image； thirdly， a cross modal interactive fusion module （CMIFM） was designed， which interactively fuse image feature information of different modalities. In order to verify the effectiveness of the proposed model， the experiment was carried out on the lung tumor PET/CT medical image dataset. Compared with the best method of the other four methods， the proposed method in the average gradient， spatial frequency， structural similarity， standard deviation， peak signal-to-noise ratio， and information entropy are improved by 1.38%， 0.39%， 29.05%， 30.23%， 0.18%， 4.63% respectively. The model can highlight the information of the lesion areas， and the fused image has clear structure and rich texture details.  
        关键词：medical image;image fusion;PET/CT;coupled generative adversarial network;swin transformer   
        118
        |
        52
        |
        0
        <HTML>
        <L-PDF><WORD><Meta-XML>
        <引用本文> <批量引用> 48418876 false
        发布时间：2024-02-01
      • “声音在目标检测领域的应用取得了新进展。针对当前仅通过监测环境声进行目标定位的方法鲁棒性较低的问题，专家提出了跨级特征知识转移下的多模态自监督目标检测网络。该网络不仅设计了基于注意力融合的多教师跨级特征知识转移损失，提高了网络的学习能力，还通过定位蒸馏损失解决了定位信息的缺失问题。在多模态视听检测MAVD数据集中的实验结果显示，该网络的mAP值在IOU值为0.5、0.75和平均的情况下较基线网络分别有6.71%、14.36%和10.32%的提升，证明了该检测网络的优越性。这一研究成果为声音在目标检测领域的应用提供了新的解决方案，也为相关领域的研究开辟了新的方向。”
        ”
        LIU Shibei,CHEN Ying
        Vol. 32, Issue 2, Pages: 237-251(2024) DOI: 10.37188/OPE.20243202.0237
        摘要：As one of the inherent properties of objects， sound can provide valuable information for target detection. At present， the method of target positioning only by monitoring environmental sound is less robust. To solve this problem， a multi-modal self-supervised target detection network under cross-level feature knowledge transfer was proposed. First of all， in view of the teachers network and students at the same characteristics of network learning ability of the limited problem， design based on the integration of teachers across level knowledge transfer loss， through the way of attention fusion deep and shallow characteristics of students， more efficient learning to the corresponding teacher middle layer characteristics， to extract more knowledge， combined with KL divergence， realize the alignment of teachers and students network alignment. In addition， in order to solve the problem of missing localization information， localization distillation loss was added， and more localization information was obtained by fitting the distribution of the teacher. With the network trained in the multimodal audiovisual detection MAVD dataset， the mAP values improve by 6.71%， 14.36% and 10.32% from the baseline network at IOU values of 0.5，0.75 and average， respectively. The experimental results demonstrate the superiority of this detection network.  
        关键词：multimodal;knowledge distillation;object detection;self-supervised;deep learning   
        88
        |
        74
        |
        0
        <HTML>
        <L-PDF><WORD><Meta-XML>
        <引用本文> <批量引用> 48418029 false
        发布时间：2024-02-01
      • “针对多模态医学图像融合中存在的纹理细节模糊和对比度低的问题，一项创新性的研究取得了显著成果。研究团队提出了一种结构功能交叉神经网络的多模态医学图像融合方法，为医学影像分析领域带来了新的突破。该方法首先设计了结构功能交叉神经网络模型，有效提取了医学图像的结构和功能信息，实现了两种信息之间的交互，从而能够精准捕捉图像的纹理细节。其次，研究团队构建了一种新型注意力机制，通过动态调整结构信息和功能信息的权重，显著提升了融合图像的对比度和轮廓清晰度。最后，他们还设计了一个从融合图像到源图像的分解过程，确保了融合图像包含更丰富的细节信息。与近年来提出的七种高水平方法相比，该方法在AG、EN、SF、MI、QAB/F和CC等客观评价指标上平均提升了22.87%、19.64%、23.02%、12.70%、6.79%和30.35%。这一成果不仅展示了该方法在纹理细节和对比度方面的优势，还证明了其在主观视觉和客观指标上的优越性。这一研究的成功为医学影像分析领域带来了新的视角和解决方案，有望为医生提供更准确、更全面的诊断依据，助力医学领域实现更精准的诊断和治疗。”
        ”
        DI Jing,GUO Wenqing,REN Li,YANG Yan,LIAN Jing
        Vol. 32, Issue 2, Pages: 252-267(2024) DOI: 10.37188/OPE.20243202.0252
        摘要：To solve the problems of texture detail blurring and low contrast in multimodal medical image fusion， a multimodal medical image fusion method with structural-functional crossed neural networks was proposed. Firstly， this method designed a structural and functional cross neural network model based on the structural and functional information of medical images. Within each structural-functional cross module， a residual network model was also incorporated. This approach not only effectively extracted the structural and functional information from anatomical and physiological medical images but also facilitated interaction between structural and functional information. As a result， it effectively captured texture details from multi-source medical images， creating fused images that closely align with human visual characteristics. Secondly， a new attention mechanism module was constructed by utilizing the effective channel attention mechanism and spatial attention mechanism model （ECA-S）， which continuously adjusted the weights of structural and functional information to fuse images， thereby improving the contrast and contour information of the fused image， and to make the fused image color more natural and realistic. Finally， a decomposition process from the fused image to the source image was designed， and since the quality of the decomposed image depends directly on the fusion result， the decomposition process could make the fused image contain more texture detail information and contour information of the source image. By comparing with seven high-level methods for medical image fusion proposed in recent years， the objective evaluation indexes of AG， EN， SF， MI， QAB/F and CC of this paper's method are improved by 22.87%， 19.64%， 23.02%， 12.70%， 6.79% and 30.35% on average， respectively， indicating that this paper's method can obtain fusion results with clearer texture details， higher contrast and better contours in subjective visual and objective indexes are better than other seven high-level contrast methods.  
        关键词：multimodal medical image fusion;structural and functional information cross-interacting network;attention mechanism;decomposition network   
        111
        |
        62
        |
        0
        <HTML>
        <L-PDF><WORD><Meta-XML>
        <引用本文> <批量引用> 48418027 false
        发布时间：2024-02-01
      • “针对遥感图像超分辨率重建领域对硬件资源的高要求，研究团队提出了一种创新的轻量级网络架构。该架构通过重参数化技术，设计出一种残差局部特征模块，有效提升了图像的局部特征提取能力。同时，引入了一个轻量级的全局上下文模块，该模块能够关联图像内部的相似特征，进而增强了网络的特征表达能力。此外，通过调整该模块的通道压缩倍数，研究团队成功减少了模型参数量，并提升了模型性能。在UC Merced遥感数据集上的测试结果显示，该算法在遥感图像3倍超分辨率下的参数量仅为539 K，远低于HSENet算法的5 470 K。同时，该算法在峰值信噪比和结构相似性方面也表现出优势，分别达到了30.01 dB和0.844 9，超过了HSENet算法的30.00 dB和0.842 0。在推理速度上，该算法也展现出明显的优势，仅需0.010 s，而HSENet算法需要0.059 s。此外，在DIV2K自然图像数据集上的测试进一步验证了该算法的泛化能力，其峰值信噪比和结构相似性相比其他算法也展现出一定的优势。这一研究成果为遥感图像超分辨率重建领域提供了新的解决方案，有望推动该领域的技术进步。”
        ”
        YI Jianbing,CHEN Junkuan,CAO Feng,LI Jun,XIE Weijia
        Vol. 32, Issue 2, Pages: 268-285(2024) DOI: 10.37188/OPE.20243202.0268
        摘要：In response to the high hardware requirements associated with the deployment of current deep learning-based remote sensing image super-resolution reconstruction models， this paper presented a lightweight， re-parameterized residual feature remote sensing image super-resolution reconstruction network. Firstly， a residual local feature module was designed using re-parameterization to effectively extract local image features. Simultaneously considering the occurrence of similar features within images， a lightweight global context module was devised to associate similar features in images， enhancing the network's feature representation capability. The channel compression rate of this module was adjusted to reduce the model's parameter count and improve its performance. Finally， a multi-level feature fusion module was employed before the upsampling module to aggregate deep features and generate a more comprehensive feature representation. Tested on the UC Merced remote sensing dataset， this algorithm exhibits a parameter count of 539 K for ×3 super-resolution， a PSNR of 30.01 dB， a SSIM of 0.844 9， and an inference time of 0.010 s. In comparison， the HSENet algorithm has a parameter count of 5 470 K， a PSNR of 30.00 dB， an SSIM of 0.842 0， and an inference time of 0.059 s. Experimental results demonstrate that this algorithm outperforms the HSENet algorithm， featuring fewer parameters， faster execution， and notable improvements in PSNR and SSIM. Testing on the DIV2K natural image dataset reveals that this algorithm exhibits advantages in PSNR and SSIM compared to other algorithms， demonstrating its strong generalization capability.  
        关键词：super resolution;remote sensing images;global context;re-parameterization;residual network   
        106
        |
        62
        |
        0
        <HTML>
        <L-PDF><WORD><Meta-XML>
        <引用本文> <批量引用> 48418025 false
        发布时间：2024-02-01
      • “混凝土表面裂缝分割技术的最新研究成果亮相。针对分割精度低、细微裂缝漏分和背景干扰等问题，研究团队提出了一种创新的联合线性引导和网格优化的裂缝分割模型。这一模型通过引入多分支线性引导模块，显著提升了网络对裂缝线性结构的表达能力，加强了不同区域裂缝间的联系，提高了全局上下文信息感知能力，进而提高了分割精度。同时，网格细节优化模块的提出，成功防止了细微裂缝的漏分，为裂缝分割领域带来了新的突破。此外，混合注意力模块的嵌入，进一步突出了裂缝特征，减少了背景干扰，进一步提升了模型的性能。在Deepcrack537、Crack500和CFD裂缝数据集上的实验结果表明，该模型的IoU值和F1-score值均明显优于大多数现有方法，展现出更高的分割精度。这一研究成果无疑为混凝土表面裂缝分割技术的发展提供了新的方向，为解决相关实际问题提供了有力支持。”
        ”
        LIU Guanghui,CHEN Jian,MENG Yuebo,XU Shengjun
        Vol. 32, Issue 2, Pages: 286-300(2024) DOI: 10.37188/OPE.20243202.0286
        摘要：A model was proposed to address issues with low segmentation accuracy， leakage of tiny cracks， and background interference in the segmentation process of concrete surface cracks. The model combined linear guidance and mesh optimization for crack segmentation. Firstly， the backbone network was enriched with a multi-branch linear guidance module. The network's ability to represent the linear structure of cracks was boosted by adaptive single-dimensional pooling. This facilitated the establishment of connections between cracks in different areas， enhanced the capability to perceive global context data， and improved the network's segmentation accuracy. Then， a module for mesh detail optimization is proposed， which divides the entire spatial domain into several spatial meshes through the three steps of partitioning， optimization， and merging. The fine cracks' information in the spatial meshes was extracted to prevent the leakage of fine cracks. Finally， a mixed attention module was embedded in the skip connections of the backbone network， highlighting crack features in the two-dimensional space and channels while also reducing background interference. On the Deepcrack537， Crack500， and CFD crack datasets， the proposed model achieves IoU values of 77.07%， 58.96%， and 56.55%， respectively. The F1-score values also performs well， achieving 87.05%， 74.19%， and 72.24%， respectively. These results are significantly better than those of most existing methods， with superior segmentation accuracy.  
        关键词：crack image;linear guide;semantic segmentation;mesh optimization;attention mechanism   
        71
        |
        47
        |
        0
        <HTML>
        <L-PDF><WORD><Meta-XML>
        <引用本文> <批量引用> 48418026 false
        发布时间：2024-02-01
      • “针对初烤烟叶等级检测领域的挑战，研究团队提出了创新的初烤烟叶等级检测网络FTGDNet。该研究旨在解决相似度较高但等级不同的初烤烟叶难以区分的问题，促进农产品精细化管理的实现。该网络通过结合CSPNet和GhostNet作为特征提取网络，显著增强了模型的特征提取能力。在主干网络末端嵌入的显式视觉中心瓶颈模块实现了全局与局部特征信息的融合，提升了特征表示的丰富性。同时，多感受野特征自适应融合模块利用注意力特征融合机制，有效融合了不同感受野的特征图，突出了有效通道信息，提高了模型的局部感受野能力。为解决定位精度下降的问题，研究团队设计了一种新的定位损失函数MCIoU_Loss，结合预测框与真实框面积损失，优化了回归定位过程。此外，引入的矩形相似度衰减系数在训练过程中动态调整了真实框与预测框的相似度判别项，加速了模型拟合。实验结果显示，FTGDNet在初烤烟叶等级检测上表现出色，验证精度高达90.0%，测试精度为87.4%，且推理时间仅为12.6 ms。与多种先进目标检测算法相比，FTGDNet在检测精度和速度上均具有明显优势，为高精度初烤烟叶等级检测提供了有力的技术支撑。这项研究不仅为初烤烟叶等级检测领域带来了新的突破，也为农产品精细化管理和智能分级设备开发提供了重要的技术支持。”
        ”
        HE Zifen,LUO Yang,ZHANG Yinhui,CHEN Guangchen,CHEN Dongdong,XU Lin
        Vol. 32, Issue 2, Pages: 301-316(2024) DOI: 10.37188/OPE.20243202.0301
        摘要：Rapid and accurate detection of flue-cured tobacco leaf grade is integral to the advancement of tobacco intelligent equipment，promoting refined management of agricultural products. Aiming at the issue that it is difficult to distinguish flue-cured tobacco leaves with high similarity between different grades， a flue-cured tobacco leaf grade detection network （FTGDNet） through multi-receptive field feature fusing adaptively and dynamic loss adjustment was proposed. Firstly， FTGDNet adopted CSPNet and GhostNet as feature extraction backbone network and auxiliary feature extraction network to enhance the model feature extraction ability， respectively；Secondly，to merge global feature information and local detail feature information，an explicit visual center bottleneck module （EVCB） was embedded at the end of backbone network； Moreover， a multi-receptive field feature adaptive fusion module （MRFA_d） was constructed， in which the attention feature fusion （AFF） mechanism adaptively fuses the weights of feature maps with different receptive fields to highlight the effective channel information while enhancing the local receptive fields of the model； In addressing the decrease of positioning accuracy due to CIoU_Loss performance degradation when the prediction box and real box shared the same aspect ratio and their centers align during the regression positioning process， a new positioning loss function MCIoU_Loss was designed， In addition， the rectangular similarity attenuation coefficient was introduced to dynamically adjust the similarity discriminant of prediction box and real box to accelerate the model fitting. The experimental results show that the verification accuracy and test accuracy of FTGDNet for 10 grades of flue-cured tobacco leaf reached 90.0% and 87.4%， respectively， with an inference time of 12.6 ms. Compared with various advanced object detection network， FTGDNet achieves higher detection accuracy and faster detection speed， which could provide technical support for high-precision flue-cured tobacco leaf grade detection.  
        关键词：flue-cured tobacco leaf;object detection;multi receptive field feature fusion;dynamic loss adjustment   
        130
        |
        70
        |
        0
        <HTML>
        <L-PDF><WORD><Meta-XML>
        <引用本文> <批量引用> 48418028 false
        发布时间：2024-02-01
      0