软分割约束边缘保持插值的半自动2D转3D

半自动2D转3D将用户标注的稀疏深度转换成稠密深度,是解决3D片源不足的主要手段之一.针对现有方法利用硬分割增强深度边缘引入误差的问题,提出像素点与超像素深度一致性约束的边缘保持插值方法.首先,建立像素点深度和超像素深度传播的能量模型,通过像素点与所属超像素间深度差异的约束项将二者关联起来;其次,利用矩阵表示形式将两个能量模型的最优化转换成一个稀疏线性方程组的求解问题.通过超像素提供的约束项,可避免深度传播穿过低对比度边缘区域,从而能保持对象边缘.实验结果表明,本文方法对象边缘处深度恢复的准确性优于融合图割的随机游走方法,PSNR改善了1.5dB以上.     

基于非局部随机游走和运动补偿的2D转3D优化方法

D转3D技术可以从2D资源中获取深度信息,以满足3D显示对3D内容的需求。针 对2D转 3D深度估计中的深度优化问题,提出一种基于非局部随机游走(NRW)和运动补偿的 深度优化算 法。本文方法在采用NRW和移动双边滤波(SBF)获得关键帧和非关键帧深度图的基础上,为 了锐化非关 键帧深度序列对象边界,结合纹理信息利用NRW算法优化深度图,同时又考虑相邻帧间的时 域信息,采 用运动补偿的方法对非关键帧深度序列进行优化,获得高质量的深度视频序列。实验结果表 明,本文方法可以得到对象边界更加准确的深度视频估计结果。     

基于间隔损失和L1范数调节的特征选择方法研究

特征评价和选择是机器学习和模式识别的重要步骤。为了获得稀疏特征子集,结合间隔损失评估策略和L1范数调节技术来获得一种有效的特征选择方法(MLFWL-L1),并将其应用到RBFSVM分类器。实验中,在UCI数据集上将提出的算法与Simba和ReliefF对比表明,验证所提出的算法是一种有效的特征选择方法。     

含有L21范数正则化的在线顺序RVFL算法

单隐层前馈神经网络(SLFN)以其量级轻、参数量少、训练成本低等优点,目前被广泛地运用于函数逼近处理、模式识别和控制领域中。随机向量函数连接网络(RVFL)作为SLFN的一种,能够将输入层与输出层做直接相连,加强输出层与输入层的关联。然而目前的预测任务中,已经训练好的网络在面对批量数据会随时间不断变化的情况时,则容易显露出泛化能力不足问题。为了提升网络的泛化能力,并防止重复训练,本文提出了一种在线顺序的RVFL算法,使用L₂₁范数实现正则化。在UCI数据集上经过对多种相关参数的最佳选择后,与同类型的RVFL算法和LR₂₁-RVFL算法相比,本文提出的LR₂₁-OSRVFL算法在多种评价指标下均有更优表现。     

LiNi1/3Co1/3Mn1/3O2粉体的类共沉淀制备及性能

在含有Li+、Co2+、Ni2+、Mn2+离子的混合溶液中加入（NH4）2CO3作沉淀剂,通过一步共沉淀反应得到含有四种金属离子的混合沉淀前驱体。前驱体经烘干,研磨后在不同温度（700～1 000 ℃）及不同时间（6～24 h）条件下进行烧结,即得到LiNi1/3Co1/3Mn1/3O2粉体。分别通过X射线衍射（XRD）、扫描电镜（SEM）及循环伏安（CV）、交流阻抗对制备粉体的微结构进行表征和对样品的电化学性能进行测试。结果表明:获得的LiNi1/3Co1/3Mn1/3O2粉体为-NaFeO2层状结构,颗粒分布均匀,放电比电容高,阻抗小。其中在900 ℃下烧结12 h所得的LiNi1/3Co1/3Mn1/3O2粉体电化学性能最优。当电压窗口在（0～1.4）Vvs.SCE、扫描速度为5 mVs-1、电解液为1 molL-1 Li2SO4溶液时,其比容量可达399.46 Fg-1;并且其阻抗也最小。     

加权SIFT流深度迁移的单幅图像2D转3D

2D视频转3D视频是解决3D片源不足的主要手段,而单幅图像的深度估计是其中的关键步骤.提出基于加权SIFT流深度迁移和能量模型优化的单幅图像深度提取方法.首先利用图像的全局描述符从深度图数据库中检索出近邻图像;其次通过SIFT流建立输入图像和近邻图像之间像素级稠密对应关系;再次由SIFT流误差计算迁移权重,将近邻图像对应像素点的深度乘以权重后迁移到输入图像上;然后利用均值滤波对迁移后的近邻图像深度进行融合;最后建立深度图优化能量模型,在尽量接近迁移后近邻图像深度的前提下,平滑梯度较小区域的深度.实验结果表明,该方法降低了估计深度图的平均相对误差,增强了深度图的均匀性.     

锂离子电池正极材料Li[Ni1/3Mn1/3Co1/3]O_2的制备与显微结构研究

本文通过正交实验探索了影响Li[Ni_1/3Mn_1/3Co_1/3]O_2材料显微结构和性能的实验因素;利用扫描电子显微镜(SEM)和透射电子显微镜(TEM)对所合成样品的微观形貌和显微结构进行表征。结果表明通过改进制备条件,可以合成颗粒尺寸细小、结构完整、结晶性好的Li[Ni_1/3Mn_1/3Co_1/3]O_2材料,这有助于改善其电化学性能。     

延时网络控制系统的鲁棒L1控制器设计

本文探讨了延时网络控制系统的鲁棒L₁控制器设计问题,针对给定的延时网络控制系统,建立了此系统的L₁(峰值-峰值)性能准则,且可以保证系统具有最劣情况下L₁性能约束且渐近稳定。在鲁棒L₁控制器中,通过锥补线性化(CCL)方法将其存在含有逆约束的LMIs形式转化为受其约束的最小化问题,这样可以运用MATLAB仿真对控制器的各参数进行求解,从而验证所提方法的可行性和有效性。     

Bi2O3对Al2O3-ZnO-Bi2O3-B2O3低熔玻璃结构和性能的影响

采用传统的熔淬技术制得了低熔Al2O3-ZnO-Bi2O3-B2O3玻璃,研究了玻璃结构、玻璃特征温度、线膨胀系数(αl)以及密度随Bi2O3含量的变化关系。结果表明:随着Bi2O3含量的增加,玻璃网络中[BO4]取代了部分[BO3],玻璃网络中出现Bi—O结构,玻璃中非桥氧的数量也逐渐增多;软化温度(ts)、玻璃化温度(tg)都是先上升后下降,而线膨胀系数先减小后逐渐增大,玻璃密度先是线性增加,然后增加趋势变大。     

Ca3(BO3)2的振动光谱分析

分别利用因子群对称分析法和位置群对称分析法对Ca₃(BO₃)₂晶体的振动模式进行了理论分析。Ca₃(BO₃)₂的晶格振动模式分为外振动和内振动模式,外振动模式为:3A_1g+4A_2g+7E_g+3A_1u+3A_2u+6E_u,内振动模式为：2A_1g+2A_2g+4E_g+2A_1u+2A_2u+4E_u。Ca₃(BO₃)₂晶体在布里渊区中心Γ点晶格振动的对称性分类为：5A_1g+6A_2g+11E_g+5A_1u+6A_2u+11E_u,其中声学模为：A_2u+E_u,拉曼活性光学模为5A_1g+11E_g,红外活性光学模为：5A_2u+10E_u,其余为非拉曼、非红外活性光学振动模。用高温固相法成功合成了Ca₃(BO₃)₂粉末,测量了它的室温Raman光谱,并利用群论分析的结果对谱图进行了讨论,指认了BO₃^3-基团的特征振动频率。     

分布式2D视频到3D视频转换系统

2D-to-3D video conversion is a feasible way to generate 3D programs for the current 3DTV industry. However, for large-scale 3D video production, current systems are no longer ade-quate in terms of the time and labor required for conversion. In this paper, we introduce a distributed 2D-to-3D video conversion system that includes a 2D-to-3D video conversion module, architecture of the parallel computation on the cloud, and 3D video cod-ing in the system. The system enables coope-ration among multiple users in the simultane-ous completion of their conversion tasks so that the conversion efficiency is greatly pro-moted. In the experiments, we evaluate the system based on criteria related to both time consumption and video coding performance.     

Polarization sensitive photodetector based on quasi-1D ZrSe3

The in-plane anisotropy of transition metal trichalcogenides (MX₃) has a significant impact on the molding of materials and MX₃ is a perfect choice for polarized photodetectors. In this study, the crystal structure, optical and optoelectronic anisotropy of one kind of quasi-one-dimensional (1D) semiconductors, ZrSe₃, are systematically investigated through experiments and theoretical studies. The ZrSe₃-based photodetector shows impressive wide spectral response from ultraviolet (UV) to near infrared (NIR) and exhibits great optoelectrical properties with photoresponsivity of 11.9 mA·W^-1 and detectivity of ~10⁶ at 532 nm. Moreover, the dichroic ratio of ZrSe₃-based polarized photodetector is around 1.1 at 808 nm. This study suggests that ZrSe₃ has potential in optoelectronic applications and polarization detectors.     

Ba(Zn_(1/3)Nb_(2/3))O_3微波介质陶瓷的研究进展

3G移动通信事业的迅速发展对应用于基站的微波介质谐振器陶瓷材料的Q值提出了更高的要求。Ba(Zn1/3Nb2/3)O3微波介质陶瓷材料因为具有很高的Q值、接近于零的τf和适宜的εr而备受关注。介绍了Ba(Zn1/3Nb2/3)O3系列微波介质陶瓷的结构和性能、改性研究、纳米化制备工艺及Ba(Zn1/3Nb2/3)O3-Ba(Co1/3Nb2/3)O3复合技术,以期对该领域其他研究者有所帮助。     

New dielectric material system of Nd(Mg1/2Ti1/2)O3–CaTiO3 at microwave frequency

The microwave dielectric properties of (1 − x)CaTiO₃–xNd(Mg_1/2Ti_1/2)O₃ (0.1  x  1.0) ceramics prepared by the conventional solid state method have been investigated. The system forms a solid solution throughout the entire compositional range. The dielectric constant decreases from 152 to 27 as x varies from 0.1 to 1.0. In the (1 − x)CaTiO₃–xNd(Mg_1/2Ti_1/2)O₃ system, the microwave dielectric properties can be effectively controlled by varying the x value. At 1400 °C, 0.1CaTiO₃–0.9Nd(Mg_1/2Ti_1/2)O₃ has a dielectric constant (ε_r) of 42, a Q × f value of 35 000 GHz and a temperature coefficient of resonant frequency (τ_f) of −10 ppm/°C. As the content of Nd(Mg_1/2Ti_1/2)O₃ increases, the highest Q × f value of 43 000 GHz for x = 0.9 is achieved at the sintering temperature 1500 °C.     

一种平面图像立体化的新方法

基于对人类心理立体视觉和生理立体视觉的理论分析和实验研究,发现在一幅平面图像的各个图像单元中引入随机交叉视差异和随机非交叉视差异之后,其双眼深度暗示效果会明显增强的立体视觉现象,并由此提出了一种新颖且实用的在仅含有心理立体视觉深度暗示的平面图像中随机加入双眼视差异的平面图像立体化方法.     

2D Molybdenum Compounds for Electrocatalytic Energy Conversion

The development of advanced nanomaterials is urgent for electrocatalytic energy conversion applications. Recently, 2D nanomaterial-derived heterogeneous electrocatalysts have shown great promise for both fundamental research and practical applications owing to their extremely high surface-to-volume ratio and tunable geometric and electronic properties. Because of their unique electronic structure and physicochemical properties, molybdenum (Mo)-based 2D nanomaterials are emerging as one of the most attractive candidates among the nonprecious materials for electrocatalysts. This review provides a comprehensive overview of the recent advances in the synthesis and modulation of 2D Mo compounds for applications in electrocatalytic energy conversion. The categories based on different compositions and corresponding synthetic approaches of 2D Mo compounds are first introduced. Subsequently, various atomic/plane/synergistic engineering strategies, along with catalytic optimization in the electrochemical process that involves the cycles of water, carbon, and nitrogen, are discussed in detail. Finally, the current challenges and future opportunities for the development of 2D Mo-based electrocatalysts are proposed with the goal of shedding light on these promising 2D nanomaterials for electrocatalytic energy conversion.