Texture branch network for chronic kidney disease screening based on ultrasound images

Chronic kidney disease(CKD) is a widespread renal disease throughout the world. Once it develops to the advanced stage, serious complications and high risk of death will follow. Hence, early screening is crucial for the treatment of CKD. Since ultrasonography has no side effects and enables radiologists to dynamically observe the morphology and pathological features of the kidney, it is commonly used for kidney examination. In this study,we propose a novel convolutional neural network(CNN) framework named the texture branch network to screen CKD based on ultrasound images. This introduces a texture branch into a typical CNN to extract and optimize texture features. The model can automatically generate texture features and deep features from input images, and use the fused information as the basis of classification. Furthermore, we train the base part of the network by means of transfer learning, and conduct experiments on a dataset with 226 ultrasound images. Experimental results demonstrate the effectiveness of the proposed approach, achieving an accuracy of 96.01% and a sensitivity of 99.44%.

Analysis of thermal conductivity in tree-like branched networks

Asymmetric tree-like branched networks are explored by geometric algorithms. Based on the network, an analysis of the thermal conductivity is presented. The relationship between effective thermal conductivity and geometric structures is obtained by using the thermal-electrical analogy technique. In all studied cases, a clear behaviour is observed, where angle （δ,θ） among parent branching extended lines, branches and parameter of the geometric structures have stronger effects on the effective thermal conductivity. When the angle δ is fixed, the optical diameter ratio β＋ is dependent on angle θ. Moreover, γand m are not related to β＊. The longer the branch is, the smaller the effective thermal conductivity will be. It is also found that when the angle θ〈δ2, the higher the iteration m is, the lower the thermal conductivity will be and it tends to zero, otherwise, it is bigger than zero. When the diameter ratio β1 〈 0.707 and angle δ is bigger, the optimal k of the perfect ratio increases with the increase of the angle δ; when β1 〉 0.707, the optimal k decreases. In addition, the effective thermal conductivity is always less than that of single channel material. The present results also show that the effective thermal conductivity of the asymmetric tree-like branched networks does not obey Murray＇s law.

THE ANALYSIS METHOD OF THE SETS OF BRANCHES BASED ON INDEPENDENT LOOPS IN THE ELECTRIC NETWORK

The matrix D describing relations of the loops to the nodes in the graph and also the setsof branches based on the independent loops and their matrix Q are defined.The theorem in whichthe product of the loop-node matrix D multiplied by the incidence matrix Aa is equal to matrix Qis put forward and proved.The admittance matrix Ylc of the sets of the branches is defined and it isassumed that the vector Vlc of voltage of the sets of branches to be a calculative quantity.The equa-tion of the sets of branches is derived and the analysis method of the sets of branches based on theindependent loops in the electric network is presented.

IMPROVED MODE-MATCHING AND NETWORK ANALYSIS OF E-PLANE WAVEGUIDE BRANCH DIRECTIONAL COUPLERS

The E-plane waveguide branch directional couplers are analyzed by a method which combines the multimode network theory with rigorous mode-matching approach. The electromagnetic field components are expanded by the superposition of LSEx modes rather than TE and TM modes in the mode-matching procedure. Meanwhile, the electromagnetic problem is transferred into the network problem through the mode-matching treatment. It is shown that the present method has the advantages of simplicity and less computation without affecting the accuracy of the calculation.

Numerical simulation of a gas pipeline network using computational fluid dynamics simulators

This article describes numerical simulation of gas pipeline network operation using high-accuracy computational fluid dynamics (CFD) simulators of the modes of gas mixture transmission through long, multi-line pipeline systems (CFD-simulator). The approach used in CFD-simulators for modeling gas mixture transmission through long, branched, multi-section pipelines is based on tailoring the full system of fluid dynamics equations to conditions of unsteady, non-isothermal processes of the gas mixture flow. Identification, in a CFD-simulator, of safe parameters for gas transmission through compressor stations amounts to finding the interior points of admissible sets described by systems of nonlinear algebraic equalities and inequalities. Such systems of equalities and inequalities comprise a formal statement of technological, design, operational and other constraints to which operation of the network equipment is subject. To illustrate the practicability of the method of numerical simulation of a gas transmission network, we compare computation results and gas flow parameters measured on-site at the gas transmission enter-prise.

International Freight Forwarding Services Network in the Yangtze River Delta, 2005–2015: Patterns and Mechanisms

This study examined the spatio-temporal trajectories of the international freight forwarding service(IFFS) in the Yangtze River Delta(YRD) and explored the driving mechanisms of the service. Based on a bipartite network projection from an IFFS firm-city data source, we mapped three IFFS networks in the YRD in 2005, 2010, and 2015. A range of statistical indicators were used to explore changes in the spatial patterns of the three networks. The underlying influence of marketization, globalization, decentralization, and integration was then explored. It was found that the connections between Shanghai and other nodal cities formed the backbones of these networks. The effects of a city's administrative level and provincial administrative borders were generally obvious. We found several specific spatial patterns associated with IFFS. For example, the four non-administrative centers of Ningbo, Suzhou, Lianyungang, and Nantong were the most connected cities and played the role of gateway cities. Furthermore, remarkable regional equalities were found regarding a city's IFFS network provision, with notable examples in the weakly connected areas of northern Jiangsu and southwestern Zhejiang. Finally, an analysis of the driving mechanisms demonstrated that IFFS network changes were highly sensitive to the influences of marketization and globalization, while regional integration played a lesser role in driving changes in IFFS networks.

Effects of refractory period on dynamical range in excitable networks

Effects of refractory period on the dynamical range in excitable networks are studied by computer simulations and theoretical analysis. The first effect is that the maximum or peak of the dynamical range appears when the largest eigenvalue of adjacent matrix is larger than one. We present a modification of the theory of the critical point by considering the correlation between excited nodes and their neighbors, which is brought by the refractory period. Our analysis provides the interpretation for the shift of the peak of the dynamical range. The effect is negligible when the average degree of the network is large. The second effect is that the dynamical range increases as the length of refractory period increases, and it is independent of the average degree. We present the mechanism of the second effect. As the refractory period increases,the saturated response decreases. This makes the bottom boundary of the dynamical range smaller and the dynamical range extend.

State Estimation of Distribution Network Considering Data Compatibility

Considering that the measurement devices of the distribution network are becoming more and more abundant, on the basis of the traditional Supervisory Control And Data Acquisition (SCADA) measurement system, Phasor measurement unit (PMU) devices are also gradually applied to the distribution network. So when estimating the state of the distribution network, the above two devices need to be used. However, because the data of different measurement systems are different, it is necessary to balance this difference so that the data of different systems can be compatible to achieve the purpose of effective utilization of the estimated power distribution state. To this end, this paper starts with three aspects of data accuracy of the two measurement systems, data time section and data refresh frequency to eliminate the differences between system data, and then considers the actual situation of the three-phase asymmetry of the distribution network. The three-phase state estimation equations are constructed by the branch current method, and finally the state estimation results are solved by the weighted least square method.

基于轻量化卷积神经网络的纬编针织物组织结构分类

为解决纬编针织物组织结构自动分类时现有方法计算量偏大的问题,基于轻量化卷积神经网络,提出了一种改进的纬编针织物组织结构分类方法。采集纬编针织物组织双面的图像,以准确判断其结构类型。在特征提取步骤中,引入了注意力机制模块,修正各个层次特征在通道域和空间域的权重。构建的双分支网络架构能并行提取织物双面的特征信息。在分类阶段,采用了串行策略来融合高维特征向量,以确定纬编针织物组织所属类别。使用准确率、宏精确率、宏召回率以及宏F_(1)评估模型的性能,并统计了参数量和计算复杂度衡量模型的资源消耗。实验结果显示,对于纬编针织物特殊的结构特点,双分支网络架构具有很好的适应性。改进后的模型增强了不同组织间的特征区分度,在受到角度旋转、尺度改变、光照条件变化等干扰下,本文方法的分类准确率可达99.51%,且保持了较小的资源消耗。

一种基于双分支注意力神经网络的皮肤癌检测框架

皮肤癌是一种主要的癌症,在过去几十年中快速增长,早期发现可以极大提高治愈率。近年来,基于皮肤镜图像利用深度学习模型(尤其是各种卷积神经网络)对皮肤癌进行识别和分类获得了广泛应用。但是与传统的图像识别分类不同,皮肤病检测任务存在数据不平衡、类间差异性小以及皮损面积占比少等方面的挑战。为此,本研究提出一种基于双分支注意力卷积神经网络(DACNN)皮肤癌分类框架。在数据预处理阶段,根据更细粒度的皮肤病类别,对数据集进行分解,降低数据不平衡程度。从网络结构上,上分支网络利用注意力残差学习(ARL)模块有效提取潜在的病变区域特征,接着利用损伤定位网络(LLN)模块定位病变区域。对其裁剪放大输入由ARL构成的下分支网络,进行局部细节的特征提取,然后结合上下分支网络的特征,进行有效的识别。最后,为了进一步缓解数据不平衡问题,在训练阶段中采用加权损失函数。在包含10015张皮肤镜图像数据集上,对所提出的DACNN模型与几种典型的皮肤病变检测框架进行了实验验证和比较。结果表明,DACNN皮肤癌变检测框架的Sensitivity、Accuracy和F1_score等性能指标分别达到了0.922、0.942和0.933,与已有的递归注意力卷积神经网络模型RACNN相比,以上3个指标分别提升了3.48%、2.95%和3.44%。总之,对于各类图像数不平衡,类间图像差异性小以及皮损面积占比少的皮肤镜图像而言,采用适当的类分解,以及双分支注意力神经网络结构首先对潜在的病变区域进行定位放大,然后进行局部细节的特征提取,能够极大的提高皮肤癌的检测准确度。

双分支结构的多层级三维点云补全

为了缓解现有点云补全方法在特征提取过程中很难平衡局部特征和全局特征的问题,提出了一种双分支结构的多层级点云补全算法。利用两个独立的分支网络分别提取出输入点云的局部特征信息和全局特征信息,再将两种特征信息进行拼接形成特征向量。使用五层联合感知机将特征向量映射成多个维度,进而提取多维特征信息并将其整合成最终特征向量。采用金字塔结构在256、512、1024特征维度上对最终特征向量进行特征解码,预测三种不同分辨率的点云。引入鉴别器网络,通过联合训练鉴别器产生的对抗损失和分层重建点云产生的补全损失去优化网络。在ShapeNet数据集上进行实验,算法显著提升了点云补全精度,并且在缺失大面积点云时也能恢复出较为完善的物体形状。

基于任务解耦的低照度图像增强方法

低照度条件下拍摄的照片往往存在亮度低、颜色失真、噪声高、细节退化等多重耦合问题,因此低照度图像增强是一个具有挑战性的任务.现有基于深度学习的低照度图像增强方法通常聚焦于对亮度和色彩的提升,导致增强图像中仍然存在噪声等缺陷.针对上述问题,本文提出了一种基于任务解耦的低照度图像增强方法,根据低照度图像增强任务对高层和低层特征的不同需求,将该任务解耦为亮度与色彩增强和细节重构两组任务,进而构建双分支低照度图像增强网络模型(Two-Branch Low-light Image Enhancement Network,TBLIEN).其中,亮度与色彩增强分支采用带全局特征的U-Net结构,提取深层语义信息改善亮度与色彩;细节重构分支采用保持原始分辨率的全卷积网络实现细节复原和噪声去除.此外,在细节重构分支中,本文提出一种半双重注意力残差模块,能在保留上下文特征的同时通过空间和通道注意力强化特征,从而实现更精细的细节重构.在合成和真实数据集上的广泛实验表明,本文模型的性能超越了当前先进的低照度图像增强方法,并具有更好的泛化能力,且可适用于水下图像增强等其他图像增强任务.

融合形状结构恢复和细节补偿的双分支点云修复网络

针对传统点云修复中难以有效地保持原始形状细节结构信息的问题,提出一种融合形状全局结构恢复和局部细节补偿的双分支点云形状修复网络.网络中的形状全局结构恢复分支为编解码-解码器结构,编码器对缺失点云数据进行特征变换以克服点云形状的旋转不变性,利用最大池化操作解决点云的无序性问题,并通过多层感知器生成原始点云的特征码字,解码器对编码得到的特征码字使用4个二维网格进行2次折叠操作,拟合点云形状得到粗修复结果;为了补偿点云粗修复结果的形状细节信息,网络中的局部细节补偿分支对编码器提取得到的不同维度特征,通过层次特征学习和多层次特征融合学习点云形状的几何结构特征,有效地恢复缺失点云数据并保留原始形状细节信息;最终将经全局结构恢复分支和局部细节补偿分支分别得到的点云数据拼接融合,再进行迭代最远点重采样,得到点云形状精修复结果.实验结果表明,在ShapeNet数据集上,所提网络比已有网络修复结果的平均CD误差和平均EMD误差分别低16%~29%和19%~65%;在ModelNet数据集上,比已有网络修复结果的平均CD误差和平均EMD误差分别低6%~41%和31%~59%;该网络可以修复原始形状的整体结构信息并能有效地恢复其形状细节,生成采样点分布均匀的完整点云模型,且对模型噪声和不同程度的模型缺失均具有鲁棒性.

基于双支路卷积网络的步态识别方法

针对步态识别易受拍摄视角、外观变化等影响的问题,提出一种基于双支路卷积网络的步态识别方法。首先,提出随机裁剪随机遮挡的数据增强方法RRDA(Restricted Random Data Augmentation),以扩展外观变化的数据样本,提高模型遮挡的鲁棒性;其次,采用结合注意力机制的两路复合卷积层(C-Conv)提取步态特征,一个分支通过水平金字塔映射(HPM)提取行人外观全局和最具辨识度的信息;另一分支通过多个并行的微动作捕捉模块(MCM)提取短时间的步态时空信息;最后,将两个分支的特征信息相加融合,再通过全连接层实现步态识别。基于平衡样本特征的区分能力和模型的收敛性构造联合损失函数,以加速模型的收敛。在CASIA-B步态数据集上进行实验,所提方法在3种行走状态下的平均识别率分别达到97.40%、93.67%和81.19%,均高于GaitSet方法、CapsNet方法、双流步态方法和GaitPart方法;在正常行走状态下比GaitSet方法的识别准确率提升了1.30个百分点,在携带背包状态下提升了2.87个百分点,在穿着外套状态下提升了10.89个百分点。实验结果表明,所提方法是可行、有效的。

基于Wi-Fi感知的多用户身份识别研究

随着无线感知技术的发展,基于Wi-Fi的身份识别研究在人机交互和家居安防等领域备受关注。尽管基于Wi-Fi信号的身份识别已经取得了初步的成功,但是目前主要适用于用户独立行为场景,并发行为下的多用户身份识别仍然面临着一系列挑战,包括用户之间的相互干扰以及模型鲁棒性差等问题。因此,提出了一种并发行为下多用户身份识别系统Wiblack,其核心思想是训练一个多分支深度神经网络(Wiblack-Net)来提取每个单用户的独特特征。首先,利用主干网络提取多用户之间的共同特征;然后,为每个用户分配一个二分类器以此判断给定群体中是否存在目标用户,在此基础上基于并发行为实现多个用户身份识别。此外,将Wiblack与多个独立的二分类模型和单个多分类模型进行对比实验,对运行效率和系统性能进行分析。实验结果显示,在同时识别3个用户身份时,Wibalck平均准确率达到了92.97%,平均精确度为93.71%,平均召回率为93.24%,平均F1值为92.43%。

基于双支特征联合映射的端到端图像去雾算法

针对卷积神经网络去雾算法中模型复杂度高、特征提取性能差等问题,本文提出了一种基于双支特征联合映射的端到端图像去雾算法.首先对大气散射模型进行变形转换,分离出模型中的双支特征;然后根据双支特点设计了两个特征提取网络MPFEM和SPFEM,分别使用两种注意力机制对其输出特征进行加权;最后将提取到的双支特征输入复原模块恢复清晰图像,并对其进行色彩增强得到最终复原效果.在模型训练过程中为避免使用单一损失函数导致纹理细节丢失等问题,采用多尺度结构相似度和平均绝对误差加权作为损失函数.实验表明,本文所提算法网络结构简单,去雾效果明显,复原图像色彩亮度保真,边缘保持性强.

多级暂堵诱导分支缝开启规律

缝内多级暂堵是提高缝内净压力,激活侧向分支缝,增大侧向改造带宽,提升层内油气动用程度的有效手段之一。研究缝内多级暂堵转向规律对指导优化暂堵方案至关重要,其难点是建立缝内暂堵模拟方法。针对裂缝性超低渗储层,基于等效黏度方法,建立了缝内多级暂堵数值模型,系统研究了逼近角、应力差、暂堵次数对分支缝开启的影响规律,研究结果表明:当水平应力差≥5 MPa且逼近角较大(≥60°)时,水力裂缝倾向于穿过天然裂缝,沿当前方向扩展,无法激活天然裂缝;水平应力差≤5 MPa时且逼近角较小(≤30°)时,水力裂缝遇到天然裂缝后能够开启钝角分支;缝内初次暂堵能够有效开启高逼近角下一级分支缝,缝内二次暂堵能够开启二级分支缝;采用缝内多级暂堵能够提高单井最高日产油量至2倍,平均日产油量提升至3倍。本文研究成果对缝内多级暂堵压裂设计奠定了模型和方法基础。

数据分布不平衡的课堂参与度自动识别研究

在线课堂学习参与度自动评估是提升课堂教学效果的重要技术途径.本文使用计算机视觉特征分析技术,提出一种在线课堂学习参与度自动识别方法.首先,采用VGGFace网络和C3D网络分别对学习者的面部表情和身体姿态进行特征编码;然后,设计基于注意力机制的双层级联聚合模块,对视频片段的特征进行融合,使得参与度高度相关的帧序列获得更大的权重.由于低参与度样本的数量和高参与度样本相比非常少,课堂参与度自动识别属于类别不平衡的数据分类问题.类别高度不均衡,导致模型训练存在很大挑战.为了缓解参与度数据分布不平衡带来的影响,本文提出采用双边分支网络作为参与度识别基本的网络结构.其中,传统学习分支进行表征学习,重新平衡分支关注少数样本分类,将特征学习和分类器学习进行分别建模.在DAiSEE数据集上的实验结果表明,提出的方法有效提升了参与度自动识别性能,尤其对少数类样本的分类具有明显的性能提升.

基于双分支特征聚合网络的车辆检测算法

车辆目标检测是自动驾驶的重要环节,现有的车辆目标检测算法在特征提取方面没有充分考虑卷积神经网络(convolutional neural network,CNN)和Transformer各自的优缺点,一定程度上限制了网络的整体性能。提出了一种由CNN和Transformer组成的双分支特征聚合网络。在编码阶段,基于CNN和Transformer各自的优势,构建了双分支主干网络来提取原始图像的特征信息;通过设计的多级别空间注意力模块和双支路特征聚合模块,使两个分支间的特征信息相互引导学习;通过构建的双分支注意力模块来进一步减少深层神经网络中特征信息的丢失。在实验部分通过消融实验和对比实验进一步验证了所提算法的有效性,其相比主流的目标检测算法,在mAP(mean average precision)指标上提升了约3.5%。

含逆变型分布式电源和分支负荷的配电网自适应电流差动保护

电流差动保护具有良好的选择性和灵敏性。但是,逆变型分布式电源(inverter-interfaced distributed generator,IIDG)和分支负荷接入改变了配电网的故障特性,现有配电网的电流差动保护面临拒动或误动的风险。为解决以上问题,该文在计及P/Q控制的IIDG和分支负荷的故障等值建模基础上,从故障点、IIDG和分支负荷的相对位置出发,挖掘配电网线路两端正序电流差的幅值特征以及正序电流故障分量的相位特征。在此基础上,提出一种正序电流幅值与故障分量相角阈值配合的双比例制动系数自适应差动保护方案,并且利用幅相平面分析所提保护判据的可靠性和灵敏性。最后通过PSCAD/EMTDC验证了所提保护方案的有效性,仿真结果表明该方案适应不同故障类型、位置和过渡电阻,且不受IIDG以及分支负荷接入的影响。