基于联合神经网络学习的中文电力计量命名实体识别

随着电力计量业务的不断扩展，迫切需要由业务信息、技术知识、行业标准及其内在联系所组成的电力计量知识图谱，为电网的决策和发展提供更为全面有效的支持。命名实体识别是构建知识图谱的基础。针对电力计量领域需要，结合中文分词技术特点，基于联合学习思想，提出了一种基于联合学习的中文电力计量命名实体识别技术。该技术联合CNN-BLSTM-CRF模型与整合词典知识的分词模型，使其共享实体类别和置信度；同时将2个模型的先后计算顺序改为并行计算，减少了识别误差累积。结果表明，在不需要人工构建特征的情况下，方法的正确率、召回率、F值等均显著优于以往方法。     

基于快速神经网络架构搜索的鲁棒图像水印网络算法

为解决深度学习在图像水印算法中计算量大且模型冗余的问题，提高图像水印算法在抵抗噪声、旋转和剪裁等攻击时的鲁棒性，提出基于快速神经网络架构搜索（neural architecture search，NAS）的鲁棒图像水印网络算法。通过多项式分布学习快速神经网络架构搜索算法，在预设的搜索空间中搜索最优网络结构，进行图像水印的高效嵌入与鲁棒提取。首先，将子网络中线性连接的全卷积层设置为独立的神经单元结构，并参数化表示结构单元内节点的连接，预先设定结构单元内每个神经元操作的搜索空间；其次，在完成一个批次的数据集训练后，依据神经元操作中的被采样次数和平均损失函数值动态更新概率；最后，重新训练搜索完成的网络。水印网络模型的参数量较原始网络模型缩减了92%以上，大大缩短了模型训练时间。由于搜索得到的网络结构更为紧凑，本文算法具有较高的时间性能和较好的实验效果，在隐藏图像时，对空域信息的依赖比原始网络更少。对改进前后的2个网络进行了大量鲁棒性实验，对比发现，本文算法在CIFAR-10数据集上对抵抗椒盐噪声和旋转、移除像素行（列）等攻击优势显著；在ImageNet数据集上对抵抗椒盐高斯噪声、旋转、中值滤波、高斯滤波、JPEG压缩、裁剪等攻击优势显著，特别是对随机移除行（列）和椒盐噪声有较强的鲁棒性。     

基于Retinex理论与概率非局部均值的红外图像增强方法

针对传统红外图像增强算法中细节模糊及过度增强的问题,提出了一种基于Retinex理论与概率非局部均值相结合的红外图像增强方法.首先通过单尺度Retinex方法调整图像中过暗与过亮部分的灰度级;然后利用概率非局部均值对图像进行分解处理得到基本层与细节层,对基本层采用直方图均衡化拉伸对比度,对细节层采用非线性函数进行增强;最后,将不同层次的结果融合得到对比度与细节增强的红外图像.用该方法对多组不同场景的红外图像进行仿真实验,并将其与多种增强方法进行主、客观对比分析,结果表明所提方法在红外图像的细节及对比度增强方面都获得了更好的效果.     

砂岩方梁断裂过程区范围的实验确定方法

裂纹前端的断裂过程区是引起岩石非线性断裂及尺寸效应的主要原因。利用数字图像相关技术对砂岩开展了三点弯曲梁实验,获得观测区域高精度的全场位移和应变数据,根据断裂韧带区域水平位移和水平应变的分布特征,结合裂尖岩石颗粒变化的微观分析,提出采用裂纹尖端水平位移波动性和水平应变突变性所得到的波动系数和水平应变突变值,确定断裂过程区形状和临界尺寸的方法。结果表明:砂岩断裂过程区的形状为不规则的狭长带状区域,断裂过程区的临界长度为11~13mm,临界宽度为1.58~2.36mm。断裂过程区区域内形变在趋向裂尖时呈指数增加,但其单位区域内的形变增量呈波动状态。该方法能够更加准确判断岩石断裂过程区的范围,有助于分析岩石的非线性断裂特性。     

基于图像处理的光学谐振腔自准直研究方法

针对我国氧碘化学激光器输出能量不断提高的情况下,主要限制光束质量的因素之一是出光过程中的光腔失调,提出了一种基于图像处理的光学谐振腔自准直研究方法。该研究方法利用氦氖穿腔光通过凹腔镜中心孔,并在凸腔镜后方放置工业相机,在屏幕上方形成一个干涉同心圆环。通过借助OpenCV库中二值化、霍夫圆检测等图像处理的方法,获得干涉同心圆环的圆心坐标,再通过以圆心为中心将图像分成上下左右四个部分,通过判断这四部分亮点像素的数量差值来预估干涉同心圆环的偏移状态,进而获得空腔条件下光学谐振腔发生变化的理论判据。实验结果表明,该方法可以获得相当精准的光学谐振腔失调判据,其失调判据选取左右亮点像素差值30000,上下亮点像素差值45000,为今后失调判据与步进电机自动调腔的结合做好充足的准备。     

Hitting probabilities and the Hausdorff dimension of the inverse images of a class of anisotropic random fields

Let X = {X(t):t ∈ R~N} be an anisotropic random field with values in R~d.Under certain conditions on X,we establish upper and lower bounds on the hitting probabilities of X in terms of respectively Hausdorff measure and Bessel-Riesz capacity.We also obtain the Hausdorff dimension of its inverse image,and the Hausdorff and packing dimensions of its level sets.These results are applicable to non-linear solutions of stochastic heat equations driven by a white in time and spatially homogeneous Gaussian noise and anisotropic Guassian random fields.     

Experimental and theoretical assessment of Fe-doped indium-oxide-based dilute magnetic semiconductors

Specimens of iron-doped indium oxide (In_1-xFe_x)₂O₃ with x?=?0.015, 0.03, 0.045 and 0.06, amalgamated through a traditional solid-state reaction method followed by H₂/air sintering, were characterised using an X-ray diffractometer (XRD), a vibrating sample magnetometer (VSM), and a scanning electron microscope (SEM) to investigate their structural, magnetic and morphological properties respectively. According to XRD plots, all the specimens exhibit cubic bixbyite structures along with ancillary phases. Magnetic assessment showed that In₂O₃ has a negative susceptibility, exhibiting diamagnetic behaviour at room temperature. The doping of Fe ions induces ferromagnetic (FM) ordering, which is enhanced with increasing doping content. The strength of the magnetisation increases when the specimens are exposed to H₂ but is reduced on further air sintering. A bound magnetic polaron (BMP) model is successfully fitted to the observed FM data involving localised carriers and magnetic cations. A multivariate assessment viz. a hierarchical cluster analysis (HCA) was used to corroborate and strengthen the experimental determined magnetic properties. A homogeneous particle distribution was observed in all SEM micrographs and is validated through MATLAB-based simulation by applying a watershed segmentation algorithm. Surface plots also confirm the change in magnetic properties with increase in doping concentration.     

Digital image correlation in experimental mechanics and image registration in computer vision: Similarities,differences and complements

Digital image correlation and image registration or matching are among the most widely used techniques in the fields of experimental mechanics and computer vision, respectively. Despite their applications in separate fields, both techniques primarily involve detecting the same physical points in two or more images. In this paper, a brief technical comparison of the two techniques is reviewed, and their similarities and differences as well as complements are presented. It is shown that some concepts from the image registration or matching technique can be applied to the digital image correlation technique to substantially enhance its performance, which can help broaden the applications of digital image correlation in scientific research and engineering practice.     

Controllable tomography phase microscopy

Tomography phase microscopy (TPM) is a new microscopic method that can quantitatively yield the volumetric 3D distribution of a sample׳s refractive index (RI), which is significant for cell biology research. In this paper, a controllable TPM system is introduced. In this system a circulatory phase-shifting method and piezoelectric ceramic are used which enable the TPM system to record the 3D RI distribution at a more controllable speed, from 1 to 40 fps, than in the other TPM systems reported. The resolution of the RI distribution obtained by this controllable TPM is much better than that in images recorded by phase contrast microscopy and interference tomography microscopy. The realization of controllable TPM not only allows for the application of TPM to the measurement of kinds of RI sample, but also contributes to academic and technological support for the practical use of TPM.