一种应用于波分复用无源光网络系统中的滤波片-PIN-TIA阵列光接收组件

设计并实现了一种将密集波分复用(DWDM)薄膜滤波片与光电探测器PIN以及前置放大器TIA集成封装的光接收组件。通过创新设计单模光纤双芯插针和自聚焦透镜的结构,解决了透过密集波分复用滤波片后的反射光偏离轴线带来的耦合问题。并基于该组件实现了32个密集波分复用通道的解复用与信号检测同时完成的波分复用无源光网络(WDM-PON)系统中上行信道阵列接收模块。测试表明模块满足国际电信联盟(ITU-T)中心波长,最小插入损耗为0.3 dB,最大插入损耗为6 dB,响应速率为100 Mbit/s~1.25 Gbit/s。     

高分辨率SAR图像目标峰值特征提取

峰值特征是进行SAR图像自动目标识别的重要特征。本文提出了SAR图像目标的通用高斯峰值模型,根据该模型提出了目标峰值提取方法。利用实测数据以及蒙特卡罗仿真结果表明:通用高斯峰值模型是实际峰值的一种精确模型,基于该模型的峰值提取方法具有良好的提取精度。     

Detection and defense of DDoS attack–based on deep learning in OpenFlow‐based SDN

Distributed denial of service (DDoS) is a special form of denial of service attack. In this paper, a DDoS detection model and defense system based on deep learning in Software‐Defined Network (SDN) environment are introduced. The model can learn patterns from sequences of network traffic and trace network attack activities in a historical manner. By using the defense system based on the model, the DDoS attack traffic can be effectively cleaned in Software‐Defined Network. The experimental results demonstrate the much better performance of our model compared with conventional machine learning ways. It also reduces the degree of dependence on environment, simplifies the real‐time update of detection system, and decreases the difficulty of upgrading or changing detection strategy.     

一种电大尺寸组合体散射的快速计算方法

采用自适应多层快速多极子算法分析电大尺寸组合体的雷达散射截面。推导了组合体表面的积分方程,通过将基函数和权函数分别用不同空间位置上的点源函数展开,自适应多层快速多极子算法实现了阻抗积分的快速计算,通过采用射线传播法,远场近似和对称性计算法则使转移因子的计算效率大大提高,所有转移过程可由快速傅里叶变换计算完成。这种算法计算组合体散射时所需的计算时间和内存显著降低,且数值计算结果和实际测试结果吻合良好。     

用激光冲击波无损评估材料内部的均一性

利用脉冲激光诱导的冲击波对材料的内部组织结构进行无损评估，首次通过测量激光在铝和碳纤维／环氧复合材料中诱导的冲击波获得了材料内部的均一性信息。     

GEPON构建中小企业专用接入网的可行性探讨

中小企业(SME)是中国经济的生力军,也是各大电信运营商收入的重要贡献者.因此,构建有竞争力的宽带接入网是目前电信运营市场竞争的新焦点.但是随着SME客户需求的不断增长,传统接入解决方案存在诸多问题,无法满足要求.本文在分析需求、问题和GEPON(吉比特以太无源光网络)的技术优势的基础上,提出利用GEPON构建SME专用接入网的设想.     

反应磁控溅射制备SiOx渐变折射率红外梳状滤光片

梳状滤光片是一种特殊的非均匀光学薄膜器件,其膜层折射率渐变分布结构使它与常规均匀光学薄膜相比具有更好的光学和机械性能.利用反应磁控溅射工艺,改变沉积SiOx(0≤x≤2)膜氧化程度,获得折射率从2.74逐渐变化到1.58(λ=1550 nm)的SiOx渐变折射率薄膜材料.通过调制膜层折射率振幅和引入膜层-外部介质折射率匹配层,成功地设计并制备了具有较好光学性能的SiOx渐变折射率红外梳状滤光片光学薄膜器件.使用单一的硅溅射靶材,通过改变氧化程度获得可变折射率材料的方法,为特殊光学薄膜器件的制备提供了一种经济实用的工艺路线.     

Moisture Sensitive Smart Yarns and Textiles from Self‐Balanced Silk Fiber Muscles

Smart textiles that sense, interact, and adapt to environmental stimuli have provided exciting new opportunities for a variety of applications. However, current advances have largely remained at the research stage due to the high cost, complexity of manufacturing, and uncomfortableness of environment‐sensitive materials. In contrast, natural textile materials are more attractive for smart textiles due to their merits in terms of low cost and comfortability. Here, water fog and humidity‐driven torsional and tensile actuation of thermally set twisted, coiled, plied silk fibers, and weave textiles from these silk fibers are reported. When exposed to water fog, the torsional silk fiber provides a fully reversible torsional stroke of 547° mm^?1. Coiled‐and‐thermoset silk yarns provide a 70% contraction when the relative humidity is changed from 20% to 80%. Such an excellent actuation behavior originates from water absorption‐induced loss of hydrogen bonds within the silk proteins and the associated structural transformation, which are corroborated by atomistic and macroscopic characterization of silk and molecular dynamics simulations. With its large abundance, cost‐effectiveness, and comfortability for wearing, the silk muscles will open up additional possibilities in industrial applications, such as smart textiles and soft robotics.     

基于混合残差密集网络的电路板红外图像超分重建北大核心CSCD

红外图像普遍存在分辨率低、细节模糊和视觉效果差的问题,使其难于直接应用在PCB故障诊断系统中。针对这一问题,本文提出了一种可充分利用红外图像层次特征的混合残差密集网络超分辨率重建算法。首先,使用卷积神经网络提取原始低分辨率图像的浅层特征信息;其次,设计多路径混合残差密集连接块,进一步提取更丰富的深层特征信息;最后,引入全局特征融合与残差学习自适应的学习并整合全局特征信息,应用转置卷积上采样完成红外图像的超分辨率重建。实验结果表明,本文算法能够有效提高重建后红外图像分辨率,使细节信息得到改善、视觉效果得到提升。基于公共/自建数据集得到的重建后图像峰值信噪比和结构相似性指标分别达到42.17 dB/39.32 dB和0.9503/0.9466,优于文中列举的双三次内插法、SRCNN和ESPCN模型,重建性能得到明显提高。