CRACK IDENTIFICATION IN STEPPED CANTILEVER BEAM COMBINING WAVELET ANALYSIS WITH TRANSFORM MATRIX

This paper illustrates the crack identification method combining wavelet analysis with transform matrix. Firstly, the fundamental vibration mode was applied to wavelet analysis. The crack location was found by the peaks of the wavelet coefficients. Secondly, based on the identified crack locations, a simple transform matrix method requiring only the first two tested natural frequencies was used to further identify the crack depth. The present method can be used for crack identification in a complex structure. Numerical results of crack identification of a stepped cantilever beam show that the suggested method is feasible.

基于Transformer紧凑编码的局部近重复视频检测算法

针对现有局部近重复视频检测算法特征存储消耗大、整体查询效率低、提取特征时并未考虑近重复帧之间细微的语义差异等问题,文中提出了一种基于Transformer紧凑编码的局部近重复视频检测算法。首先,提出了一个基于Transformer的特征编码器,其学习了大量近重复帧之间细微的语义差异,可以在编码帧特征时对各个区域特征图引入自注意力机制,在有效降低帧特征维度的同时也提高了编码后特征的表示性。该特征编码器通过孪生网络训练得到,该网络不需要负样本就可以有效学习近重复帧之间的相似语义信息,因此无需沉重和困难的难负样本标注工作,使得训练过程更加简易和高效。其次,提出了一个基于视频自相似度矩阵的关键帧提取方法,可以从视频中提取丰富但不冗余的关键帧,从而使关键帧特征序列能够更全面地描述原视频内容,提升算法的性能,同时也大幅减少了存储和计算冗余关键帧带来的开销。最后,基于关键帧的低维紧凑编码特征,采用基于图网络的时间对齐算法,实现局部近重复视频片段的检测和定位。该算法在公开的局部近重复视频检测数据集VCDB上取得了优于现有算法的实验性能。

Error modeling of 3-RSR parallel robot based on D-H transformation matrix

By selecting any one limb of 3-RSR parallel robot as a research object, the paper establishes a position and orienta- tion relationship matrix between the moving platform and the base by means of Denavit-Hartenberg （D-H） transformation matrix. The error mapping model is derived from original error to the error of the platform by using matrix differential method. This model contains all geometric original errors of the robot. The nonlinear implicit function relation between po- sition and orientation error of the platform and the original geometric errors is simplified as a linear explicit function rela- tion. The results provide a basis for further studying error analysis and error compensation.

Transformer与CNN并行引导的水下图像增强

为克服水下图像对比度低和色偏的问题,提出了基于Transformer与CNN并行引导的水下图像增强算法。利用3D位置嵌入模型为Transformer提供相对位置信息、色偏信息和特征图的全局特征,利用CNN编码器提取图像局部特征,将Transformer提取的全局特征和CNN提取的局部特征通过特征调制矩阵整合在一起,通过CNN解码器提高图像的分辨率,将解码器输出的特征图输入到特征加强网络中,由特征加强网络输出最终结果。采用现有的EUVP配对数据集进行训练,为验证该算法的优越性,选取具有不同程度色偏的水下图像进行定性比较和定量实验,结果显示,该算法增强后的水下图像峰值信噪比指标(peak signal-to-noise ratio,PSNR)和结构相似性指标(structural similarity index measure,SSIM)均高于其他对比算法,主观质量也得到显著提高,能够产生颜色丰富且清晰度较高的增强图像。

Determination of Transmission Routing of OXC Nodes Based on AWG Multiplexer by Matrix Transformation

The structures of the space switching and the wavelength switching optical cross connect (OXC) nodes which are based on the arrayed waveguide grating (AWG) multiplexer are analyzed.By the matrix transformation relation between the input and output wavelengths of the AWG multiplexer, the wavelength transmission routings of the space switching and wavelength switching OXC nodes are determined.

BANACH SPACES AND INEQUALITIES ASSOCIATED WITH NEW GENERALIZATION OF CESàRO MATRIX

Let the triangle matrix A^(ru)be a generalization of the Cesàro matrix and U∈{c_(0),c,l_(∞)}.In this study,we essentially deal with the space U(A^(ru))defined by the domain of A^(ru)in the space U and give the bases,and determine the Kothe-Toeplitz,generalized K?theToeplitz and bounded-duals of the space U(A^(ru)).We characterize the classes(l_(∞)(A^(ru)):l_(∞)),(l_(∞)(A^(ru)):c),(c(A^(ru)):c),and(U:V(A^(ru)))of infinite matrices,where V denotes any given sequence space.Additionally,we also present a Steinhaus type theorem.As an another result of this study,we investigate the l_(p)-norm of the matrix A^(ru)and as a result obtaining a generalized version of Hardy's inequality,and some inclusion relations.Moreover,we compute the norm of well-known operators on the matrix domain l_(p)(A^(ru)).

Matrix transformations of l^q(X) to l^p(Y )

Let X and Y be Banach spaces, 0 〈 q 〈 ＋∞, i ≤ p 〈 ＋∞. In this paper, we characterize matrix transformations of lq （ X ） to lp （ Y ）.

结合Transformer的轻量化中文语音识别

近年来,深度神经网络模型在语音识别领域成为热门研究对象。然而,深层神经网络的构建依赖庞大的参数和计算开销,过大的模型体积也增加了其在边缘设备上部署的难度。针对上述问题,提出了基于Transformer的轻量化语音识别模型。首先使用深度可分离卷积获得音频特征信息;其次构建了双半步剩余权重前馈神经网络,即Macaron-Net结构,并引入低秩矩阵分解,实现了模型压缩;最后使用稀疏注意力机制,提升了模型的训练速度和解码速度。为了验证模型,在Aishell-1和aidatatang_200zh数据集上进行了测试。实验结果显示,与Open-Transformer相比,所提模型在字错误率上相对下降了19.8%,在实时率上相对下降了32.1%。

基于Dlib和变种Transformer的哈欠检测方法

为及时检测驾驶员的疲劳驾驶情况,提出了一种基于Dlib和变种Transformer的哈欠检测方法。首先,基于Dlib的人脸关键点模型构建驾驶员眼部和嘴部的哈欠特征矩阵,然后在视频检测领域提出一种变种Transformer模型对哈欠特征矩阵进行提取和分类,最后基于YawDD数据集进行验证,结果表明,所提出算法的哈欠检测准确率达96.8%,高于已有算法,适用于驾驶员疲劳驾驶时哈欠行为的检测。

Research on the Applications of Probability Model and Matrix Transformation in Data Analysis and Management under the Environment of Information Time

In this paper, we research the probability theory and matrix transformation based technique to manage the data for processing and analysis. Clustering analysis research has a long history, over the decades, the importance and the cross characteristics with other research direction to get the affirmation of the people. The probability theory and linear algebra act as the powerful tool for analyzing and mining data. The experimental result illustrates the effectiveness. In the near future, we plan to conduct more theoretical analysis on the topic.

Transformation Matrix for Combined Loads Applied to Thin-Walled Structures

This paper transforms combined loads, applied at an arbitrary point of a thin-walled open section beam, to the shear centre of the cross-section of the beam. Therein, a generalized transformation matrix for loads with respect to the shear centre is derived, this accounting for the bimoments that develop due to the way the combined loads are applied. This and the authors’ earlier paper (World Journal of Mechanics 2021, 11, 205-236) provide a full solution to the theory of thin-walled, open-section structures bearing combined loading. The earlier work identified arbitrary loading with the section’s area properties that are necessary to axial and shear stress calculations within the structure’s thin walls. In the previous paper attention is paid to the relevant axes of loading and to the transformations of loading required between axes for stress calculations arising from tension/compression, bending, torsion and shear. The derivation of the general transformation matrix applies to all types of loadings including, axial tensile and compression forces, transverse shear, longitudinal bending. One application, representing all these load cases, is given of a simple channel cantilever with an eccentrically located end load.

Orthogonalization of unified and extended Bézier basis and its transformation matrix

UE-Brzier （unified and extended Brzier） basis is the unified form of Brzier-like bases, including polynomial Brzier basis, trigonometric polynomial and hyperbolic polynomial Brzier basis. Similar to the original Brzier-like bases, UE-Brzier basis func-tions are not orthogonal. In this paper, a group of orthogonal basis is constructed based on UE-Brzier basis. The transformation matrices between UE-Brzier basis and the proposed orthogonal basis are also solved.

Inverse Transformation of Elliptical Relative State Transition Matrix

A new set of relative orbit elements (ROEs) is used to derive a new elliptical formation flying model in previous work. In-plane and out-of-plane relative motions can be completely decoupled, which benefits elliptical formation design. In order to study the elliptical control strategy and perturbation effects, it is necessary to derive the inverse transformation of the relative state transition matrix based on relative orbit elements. Poisson bracket theory is used to obtain the linear transformations between the two representations: the relative orbit elements and the geocentric orbital frame. In this paper, the details of these transformations are presented.

A Direct Transformation of a Matrix Spectrum

A method is presented for calculating a matrix spectrum with a given set of eigenvalues. It can be used to build systems with different spectrums with the aim of choosing desired alternative. It enables a practical implementation of control algorithms without resorting to transformation of variables.

Effects of Heparin on Transforming Growth Factor-β_1 and Extracellular Matrix Components in the Glomeruli of Diabetic Rats

The effects of heparin on the expression of transforming growth factor-β 1 (TGF-β 1) and two extracellular matrix components laminin (LN) and fibronectin (FN) in diabetic rat glomeruli were investigated. Twenty-six rats were randomly divided into control group (C, n=8), diabetic group (D, n=9), and diabetes+heparin group (DH, n=9). After 8-week therapy of heparin (200 U once daily by abdominal injection), TGF-β 1, LN and FN expression in glomeruli was detected by immunohistochemical method. The results showed that the expression levels of TGF-β 1, LN and FN were higher in group D than in group C. It was found that heparin could reduce 24-h urinary albumin excretion and inhibit overexpression of TGF-β 1, LN and FN in glomeruli of diabetic rats. It suggested that the inhibitory effect of heparin on diabetic glomerular sclerosis was at least partly related with the inhibition of TGF-β 1 expression.

基于矩阵乘积态的有限纠缠量子傅里叶变换模拟

与经典计算不同,在量子计算中量子比特可以处于叠加态,多个量子比特之间还可以形成纠缠态。表示n个量子比特组成的量子态需要存储2^(n)个振幅,这种指数级的存储开销使得大规模的量子模拟难以进行。然而当量子态的纠缠程度有限时,使用矩阵乘积态表示量子态仅需要线性的空间复杂度,可以扩大模拟的规模。使用HIP-Clang语言,基于CPU+DCU的异构编程模型,使用矩阵乘积态表示量子态,对量子傅里叶变换进行模拟。结合矩阵乘积态的特点,对量子傅里叶变换线路进行分析,减少模拟实现时不必要的张量缩并运算与正交化构建。对模拟过程中的张量缩并进行分析,使用TTGT算法完成张量缩并运算,同时利用DCU的并行处理能力来提高效率。对模拟结果进行分析,分别通过振幅误差与半经典Draper量子加法器的结果验证了模拟的正确性。对模拟规模进行分析,当量子态的纠缠熵最大时,使用16 GB的内存空间最多只能模拟24位的量子态,而当量子态内部纠缠程度较低时,可以对上百位的量子态进行量子傅里叶变换模拟。

融合角点检测的点云配准研究

为提高点云配准效率,提出一种基于点云曲率信息的Harris角点提取算法,通过引入曲率约束,对传统Harris角点提取做出改进;计算每个点的曲率值并与整幅点云的平均曲率作比较,保留曲率值大于平均曲率的点云,实现对配准价值较低的点的剔除,为后续点云配准减少计算量;构建所提取角点之间的特征匹配,结合采样一致性(sample consensus initial alignment, SAC-IA)粗配准和正态分布变换(normal distributions transform, NDT)精配准,寻找最优的变换矩阵,实现源点云和目标点云的重合。通过比较不同算法在公开数据集上进行实验验证,结果表明该方法对复杂点云、多拐点点云都有较好的表现,对初始位置相差较大的点云也有良好的适用性。

矩阵初等变换的几何意义

矩阵的初等变换在线性代数中应用十分广泛,以初等行变换为例,分别给出了非退化的初等变换、退化的初等变换的实例与对应的几何图形,在此基础上给出了任意角度的旋转.在教学过程中,通过挖掘初等变换的几何意义,可以使学生建立代数与几何的统一观,促进学生抽象思维与形象思维协调发展.

基于多平面分割和矩阵变换的航摄边坡点云滤波算法

点云滤波是实现地面点与非地面点分离,获取最真实地面点云的重要处理手段,是进行公路边坡点云灾害识别的基础.为了克服传统点云滤波算法在边坡场景下处理速度慢、结果精度低和误差大等问题,提出了基于多平面分割和矩阵变换的航摄边坡点云滤波算法.该方法首先利用基于曲率的区域生长算法对边坡进行多平面分割,得到多个边坡子点云;其次拟合得到边坡子点云平面模型,再利用旋转矩阵将边坡子点云进行空间转换至水平面;然后通过模拟布料下沉并设置距离阈值分离出非地面点;最后再次利用旋转矩阵的逆矩阵进行空间位置还原,进而得到滤波后的边坡点云.利用精细贴近仿地航线设计方法获取高精度点云模型以在不同的边坡场景下进行算法测试,并与其他传统滤波算法结果进行对比,结果表明:在所有的试验中,本文算法的总误差分别为7.11%、4.15%、1.45%、4.41%,在所有测试算法中最小;Kappa系数分别为0.77、0.90、0.96、0.90,在所有测试算法中最大.本文提出的算法在面对地形和植被情况复杂的边坡情景下,有着较高的准确性以及较强的适用性,为公路边坡点云滤波提供了新的解决方案.

基于Hatch滤波的改进矩阵变换法在BDS多频模糊度解算中的应用

长基线条件下,空间相关性弱,电离层延迟残差和噪声影响变大,致使矩阵变换法模糊度解算过程中,窄巷组合受噪声影响大,单历元模糊度解算成功率低。因此,削弱电离层残差及噪声影响可以一定程度提高模糊度解算成功率。本文在分析矩阵变换算法基础上,利用Hatch滤波有效抑制伪距测量噪声和多径效应特点,结合弱电离层、低噪声的载波组合,通过相位平滑伪距代替组合伪距值参与解算,从而削弱电离层残差及观测噪声影响。通过实测数据分析,本文算法可有效提高矩阵变换法模糊度解算成功率。