一种新型三维光正交码的设计

基于光正交码(OOC)编码技术,提出了一种新型空/时/频三维OOC的设计方案,对光信号在时、频和空域进行编码,该码具有理想的自/互相关特性。通过仿真比较,在容量和系统的误码率等性能方面比一、二维OOC都有较大的提高,因而能很好地适应未来高密集光纤通信。     

多分辨率融合输入的U型视网膜血管分割算法

针对视网膜血管拓扑结构不规则、形态复杂和尺度变化多样的特点,该文提出一种多分辨率融合输入的U型网络(MFIU-Net),旨在实现视网膜血管精准分割。设计以多分辨率融合输入为主干的粗略分割网络,生成高分辨率特征。采用改进的ResNeSt代替传统卷积,优化血管分割边界特征;将并行空间激活模块嵌入其中,捕获更多的语义和空间信息。构架另一U型精细分割网络,提高模型的微观表示和识别能力。一是底层采用多尺度密集特征金字塔模块提取血管的多尺度特征信息。二是利用特征自适应模块增强粗、细网络之间的特征融合,抑制不相关的背景噪声。三是设计面向细节的双重损失函数融合,以引导网络专注于学习特征。在眼底数据用于血管提取的数字视网膜图像(DRIVE)、视网膜结构分析(STARE)和儿童心脏与健康研究(CHASE_DB1)上进行实验,其准确率分别为97.00%, 97.47%和97.48%,灵敏度分别为82.73%, 82.86%和83.24%,曲线下的面积(AUC)值分别为98.74%, 98.90%和98.93%。其模型整体性能优于现有算法。     

WDM光网络关键技术研究

信息爆炸导致了对带宽的无限需要,而波分复用（ＷＤＭ）系统在满足不断增长的带宽需求的同时还具有不可比拟的联网优势,文中着重研究了ＷＤＭ光网络的关键技术。     

正交读出方式体全息光栅通信波长衍射特性

为了寻求高质量和高密度的密集波分复用器件,采用了在双掺铟铁的铌酸锂晶体中透射式记录/正交式读出方案制作体全息光栅的方法,对体全息光栅衍射特性进行理论分析和实验验证.利用波长为532nm的激光记录尺寸比为1:1的体全息光栅,然后用中心波长为1550nm的红外通讯波长成功读出,取得了波长选择性为0.5nm的波长衍射特性数据.同时,利用2维耦合波理论的闭形式解析解得到了该体全息光栅衍射效率随波长的变化关系.结果表明,实验结果与理论预期相符合,这一方法对制作体全息光栅密集波分复用器件的实用化是有帮助的.     

GPU集群能耗优化控制模型研究

随着大数据技术的发展,GPU集群作为一种高效的并行系统被应用到大规模数据实时计算中.能量是实时计算时重要的资源,GPU集群的能耗优化及实时消减成为一个具有挑战性的问题.从集群全局角度引入模型预测控制策略,并建立闭环反馈机制的多输入多输出控制器.通过调整计算频率和改变活跃流多处理器来改变能耗状态,利用反馈和滚动优化机制完成对未来的控制预判,实现消减冗余能耗的目标.实验表明:控制模型的精度和节能效果优于基准模型,而且具有较好的稳定性,适合应用到大规模数据实时计算中.     

基于富铟团簇量子结构的双波长可调谐半导体激光器

本研究提出了一种采用增益芯片和光栅外腔的双波长可调谐半导体激光器。增益芯片采用了富铟团簇量子限制结构作为其量子限制结构,由于该结构独特的平顶增益特性,可以使激光器在双波长调谐范围内实现光强稳定。本研究中的谐振腔包括内部谐振腔和外部谐振腔,其中内部谐振腔由增益芯片的两个自然解理面构成,以支撑整个系统工作在特定波长。外部谐振腔由增益芯片的一个自然解理面和光栅构成,可实现969.1~977.9 nm的工作波长范围。最终该系统基于单个增益芯片和单个光栅实现了同步双波长输出,双波长的频率差在THz范围。本研究有望为实现双波长差频太赫兹源提供一种可能的解决方案。     

Highly Dispersed NiO Clusters Induced Electron Delocalization of Ni?N?C Catalysts for Enhanced CO2 Electroreduction

Oxygen-regulated Ni-based single-atom catalysts (SACs) show great potential in accelerating the kinetics of electrocatalytic CO₂ reduction reaction (CO₂RR). However, it remains a challenge to precisely control the coordination environment of Ni O moieties and achieve high activity at high overpotentials. Herein, a facile carbonization coupled oxidation strategy is developed to mass produce NiO clusters-decorated Ni N C SACs that exhibit a high Faradaic efficiency of CO (maximum of 96.5%) over a wide potential range (−0.9 to −1.3 V versus reversible hydrogen electrode) and a high turnover frequency for CO production of 10 120 h⁻¹ even at the high overpotential of 1.19 V. Density functional theory calculations reveal that the highly dispersed NiO clusters induce electron delocalization of active sites and reduce the energy barriers for *COOH intermediates formation from CO₂, leading to an enhanced reaction kinetics for CO production. This study opens a new universal pathway for the construction of oxygen-regulated metal-based SACs for various catalytic applications.     

Hybrid Catalyst Coupling Zn Single Atoms and CuNx Clusters for Synergetic Catalytic Reduction of CO2

Reverse water-gas shift (RWGS) reaction is the initial and necessary step of CO₂ hydrogenation to high value-added products, and regulating the selectivity of CO is still a fundamental challenge. In the present study, an efficient catalyst (CuZnN_x@C-N) composed by Zn single atoms and Cu clusters stabilized by nitrogen sites is reported. It contains saturated four-coordinate Zn-N₄ sites and low valence CuN_x clusters. Monodisperse Zn induces the aggregation of pyridinic N to form Zn-N₄ and N₄ structures, which show strong Lewis basicity and has strong adsorption for *CO₂ and *COOH intermediates, but weak adsorption for *CO, thus greatly improves the CO₂ conversion and CO selectivity. The catalyst calcined at 700 °C exhibits the highest CO₂ conversion of 43.6% under atmospheric pressure, which is 18.33 times of Cu-ZnO and close to the thermodynamic equilibrium conversion rate (49.9%) of CO₂. In the catalytic process, CuN_x not only adsorbs and activates H₂, but also cooperates with the adjacent Zn-N₄ and N₄ structures to jointly activate CO₂ molecules and further promotes the hydrogenation of CO₂. This synergistic mechanism will provide new insights for developing efficient hydrogenation catalysts.     

传感器网络中基于二元簇的节能路由算法

传感器网络中的分簇算法已经被证明能够有效地节约能量。据此提出了一种基于二元簇的路由算法,其基本思想是网络中的每个节点同时属于2个不同簇,利用二元性来减少在路由发现中的资源浪费,并能够更加确定地发现路由以节约资源。理论分析和模拟实验都表明,该算法在能量消耗和网络生命周期延长方面较优。