基于小波变换和均值滤波的图像去噪方法

将小波变换和均值滤波相结合提出了一种有效的图像去噪方法,先将含噪图像进行小波分解,获得不同频带的子图像.将低频近似图像保持不变,对水平、垂直和对角三个方向高频细节图像根据其特性采用三种不同形状的模板进行均值滤波,最后将低频近似图像与三个均值滤波后高频细节图像合成得到去噪后的图像.实验结果表明,该方法在降低了图像噪声的同时又尽可能地保留图像的细节,其去噪效果优于单一小波阈值法和均值滤波法.     

基于曲率约束的点云分割去噪方法

在点云去噪处理过程中,为提高对曲率变化较大区域的去噪效果,提出基于曲率约束的点云分割去噪方法。该方法通过曲率约束点云数据使模型特征得到有效保持,构建噪声光顺的映射函数使得噪声点回归光顺,能够避免使用传统滤波对点云数据模型产生过光顺,对后续处理奠定一定基础。实验表明,该方法相对于双边滤波能够有效地保持模型的特征,保留模型边缘信息,去除噪声效果更为明显。     

基于采样保真性的点模型去噪

提出了一种基于采样保真性的点模型去噪算法.该算法通过移动最小二乘曲面,计算每个采样点的保真性;由法向张量投票方法,测量采样点的特征性;利用改进的双边滤波算子获得各个采样点的滤波方向,结合保真性和特征性对点模型去噪.实验结果表明,算法是鲁棒的,在剔除噪声的同时能够有效地保持曲面的几何特征.     

图像检测方法研究

图像噪声会严重影响其视觉效果,采用适当的方法进行图像去噪是一项必不可少的预处理操作。本文对几种常用的图像去噪方法进行了分析,通过仿真高斯噪声和椒盐噪声,着重对基于均值滤波、中值滤波以及卷积神经网络的实际去噪效果进行了比较。实验结果表明三种方法的去噪效果各有优劣,均能提高图像的检测质量。     

数字射线图像的滤噪方法

目的 研究射线图像中保持细节的滤噪方法。方法 依据图像中噪声的特点,提出三种去噪方法,即差分梯度法、邻域比较法、σ滤波与邻域比较结合。结果 采用这些方法对射线图像进行处理,有明显的去噪效果,且保持细节。结论这些方法对于图像滤噪具有普遍意义,特别对于含有微小细节、颗粒脉冲噪声的图像更适合。     

数字射线图像的滤噪方法

目的　研究射线图像中保持细节的滤噪方法 .方法　依据图像中噪声的特点 ,提出三种去噪方法 ,即差分梯度法、邻域比较法、σ滤波与邻域比较相结合 .结果　采用这些方法对射线图像进行处理 ,有明显的去噪效果 ,且保持细节 .结论　这些方法对于图像滤噪具有普遍意义 ,特别对于含有微小细节、颗粒脉冲噪声的图像更适合     

基于图像去噪的混合噪声滤波算法

传统的中值滤波和均值滤波通常被分别用来滤除椒盐噪声和高斯噪声.但是当图像同时存在高斯噪声和椒盐噪声时,单独使用哪种滤波方法都不会达到最好的去噪效果.为了能同时滤除两种不同性质的噪声,提出了一种新的自适应混合噪声滤波算法.该算法采用了一种基于自适应阈值的方法对滤波系数加以优化,使其既可以有效地减少噪声,又可以较好的保持图像的边缘细节信息,仿真结果表明该算法能较好的滤除混合噪声,且滤波效果优于传统的滤波算法.     

基于曲率特征混合分类的高密度点云去噪方法

为解决复杂曲面点云在平滑去噪中存在的问题,提出基于曲率信息混合分类的特征保持点云平滑算法．该方法将平面投影与双边滤波算法相结合,采用主成分分析法对点云的局部曲率特性进行评价,使用线性组合混合分类方法将数据分为平面、次特征、富特征类型以及组合类型．针对不同特征邻域类型,提出平面类型的投影平滑方法、次特征和富特征类型的变参数双边滤波法平滑方法的线性组合方法实现点云数据的平滑去噪．将该方法用于激光三维高分辨率人体扫描系统所得到的高密度点云数据,实验结果表明该方法能够在有效光顺点云的同时保持其表面的几何特征,且简化了法向调整的繁杂运算．     

基于双边滤波的自适应彩色图像去噪研究

目的为了克服彩色图像去噪后存在的特征模糊,研究基于双边滤波的自适应彩色噪声图像去噪方法。方法使用二维离散小波变换(DWT)对含噪声的彩图图像进行近似分量、水平细节分量、垂直细节分量和对角细节分量等4个方向的分解。根据DWT各方向分量归一化后的方差比例,利用RBF神经网络构造双边滤波系数模型确定不同方向的最佳去噪系数,提出彩色噪声图像自适应去噪方法(DWT-ABF),并将该方法与常规方法作对比。结果在不同噪声类型以及混合噪声失真情况下文中方法都能有效地去除噪声,并同时保留图像细节信息,且与其他方法相比,文中方法去噪后的图像都具有更高的PSNR值。结论文中方法克服了传统双边滤波无法自行确定最佳参数的缺陷,同时也良好地解决了去噪图像特征模糊的问题。     

高速X射线线阵探测器噪声分析与滤波方法

对高速X射线线阵探测器的噪声进行了分析,将探测器噪声分为本底噪声、随机噪声、孤立噪点三种类型。通过实验,研究了随着探测器扫描速度增加,噪声的变化情况;分析了扫描速度越快噪声对探测器探测结果影响越大的原因。提出了一种适用于高速X射线线阵探测器的滤波方法:采用差值法滤除本底噪声;采用卡尔曼滤波器滤除随机噪声;采用根据噪声特点改进的中值滤波法滤除孤立噪点。实验证明,该方法既可以有效滤除探测器各类噪声,又能很好的保持探测器探测图像的细节。     

基于向量夹角的三角网格模型简化算法

提出以顶点向量夹角为剔除准则,运用顶点去除方法进行网格简化的算法。该算法首先计算每个网格顶点与其周围相邻顶点所形成的方向向量,然后根据方向向量的平均夹角来识别和简化网格中的平面区域,之后再以该顶点所处区域是否比较平坦为顶点去除限制条件,判断该顶点是否符合安全去除原则,这样可以保留特征顶点,进而保持物体的视觉特征。在顶点去除后,还需要对因此而形成的多边形进行三角形网格重建,以填补空洞。该算法在简化质量和简化速率方面取得了较好的平衡。     

六面体网格体积成形有限元分析关键技术

为提高三维体积成形有限元分析软件的计算精度、减少网格数量,基于六面体网格特性研究了六面体网格体积成形有限元分析系统的关键技术,通过对STL文件的索引重构建立了模具实体模型顶点和三角形面片拓扑连接关系,提出了一种局部坐标系建立方法保证触模和对称约束条件的正确施加,给出了相对速度和相对位置两种节点触模判断方法,并提出了调整触模节点位置的最短距离法,与初矢修正法和原长修正法比较,其调整距离短,体积损失小.基于以上技术,开发了基于三维六面体网格的体积成形有限元分析系统,对典型体积成形工艺进行了有限元模拟,并与Deform模拟结果进行对比,二者吻合较好,验证了所建立模型与相应处理技术的可行性.     

法矢量法实现网格模型简化

提出一种基于顶点法矢量和面片法矢量的网格简化算法,能在大幅度简化的情况下,简化模型依然能保持良好的视觉效果.算法采用边折叠实现网格模型的简化,首先,建立网格模型上每个顶点的不平度,以此来衡量顶点局部对视觉效果的贡献程度;其次,度量三角形在边折叠后的变形误差,用于衡量边折叠对视觉效果所造成的畸变程度;最后,综合顶点不平度和三角形的变形误差,建立边折叠代价函数,并以此指导网格的简化.此外,在此简化算法的基础上,还提出一个递进网格传输的框架,并实现了一个基于浏览器的可视化原型系统.     

A parallel mesh generation algorithm based on the vertex label assignment scheme

In this article a new mesh generation algorithm is presented. The algorithm is based on a new approach called the vertex label assignment scheme to provide the information for the mesh generation so that parallel processing becomes possible. The algorithm generates 2D meshes of quadrilaterals on the basis of individual faces; conformity and smoothness of the resultant mesh are automatically assured. Local and selective mesh-refinements are also supported. A regular quadrilateral network which defines the geometry of the problem and an associated subdivision level assignment which specifies mesh density data on the network are the only input information.     

Multiphase hybrid stress finite element analysis of heterogeneous media by simple mesh: One element with one interface

In this paper, a finite element method is proposed to analyze the microscopic and macroscopic mechanical behaviors of heterogeneous media with randomly distributed inclusions. A simple mesh partitioned the domain into regular quadrilateral or triangular elements, where one element may contain two phases. An assumed stress hybrid formulation is implemented in the finite element model and the functional is derived for an element containing two phases. Numerical examples were used to study the microscopic and macroscopic properties of the composites, such as the effective modulus, to validate of the proposed model. The results show that the proposed multiphase hybrid stress finite element model can accurately measure the stress fields of materials with arbitrary microstructural distributions and improve computational efficiency by about 30 to 1500 times in comparison with the traditional displacement based finite element method.     

A displacement scheme with drilling degrees of freedom for plane elements

A scheme of boundary displacements with drilling degrees of freedom for plane elements is presented. The scheme is free from zero displacement modes and allows the development of hybrid finite elements with vertex and mid-side nodes, each node including a drilling degree of freedom besides the translational ones. Four quadrilateral isoparametric hybrid stress elements are implemented, and numerical results for some current test problems are given.     

基于SDAE和GRUNN的行星齿轮故障识别

针对噪声环境和时变转速工况下行星齿轮故障识别率低的问题,提出一种基于堆叠消噪自动编码器(SDAE)和门控循环单元神经网络(GRUNN)的行星齿轮故障识别方法.构建基于SDAE和GRUNN的混合模型,处理前后关联的时序数据,自动地从含噪样本中提取鲁棒故障特征;将行星齿轮故障诊断的训练样本看作该混合模型的输入数据,采用Ad...     

Numerical simulation of quasibrittle fracture in concrete

A triangular unit, constructed from constant strain triangles with nodes along its sides and not at the vertex, is developed for the simulation of fracture in quasibrittle materials. Fracture is modelled through a constitutive softening-fracture law at the interface nodes, with the material within the triangular unit remaining linear elastic. The inelastic displacement at an interface node represents the crack opening, which is related to the conjugate inter-nodal force by the appropriate softening relationship. The path-dependent softening behaviour is solved in nonholonomic rate form within a quasiprescribed displacement formulation. At each event in the loading history, all equilibrium solutions for the prescribed mesh can be established and the critical equilibrium path with the minimum increment of external work adopted. The crack profile or trajectory is restricted in that it can only follow the interface boundaries of the defined mesh. No remeshing is carried out. Solutions to the nonholonomic rate formulation are obtained using a mathematical programming procedure based on the solution of an LCP. Several examples are given and compared, where possible, with published results. The advantage of this formulation is that branching and interacting cracks can be tracked subject to the limitations of the prescribed mesh.     

A Novel Hybrid Element Analysis for Piezoelectric-parent Material Wedges

A new super wedge-tip hybrid element together with standard hybrid finite elements is developed to determine singular electro-elastic fields at the vertex of piezoelectric-parent material wedges. With the technique, stress and electric displacement intensity factors and energy release rates in a PZT5H panel containing a central crack are computed and compared with the available theoretical solutions. It is shown that the numerical results converge to exact solutions rapidly with fewer elements and proper number of high order terms. Then, the fracture parameters named as general stress intensity parameters and general electric displacement intensity parameters for three kinds of bimaterial wedges such as bi-piezoelectric wedge, piezoelectric-conductor wedge and piezoelectric-composite wedge are computed and depicted in graphical forms.     

Coarsening unstructured meshes by edge contraction

A new unstructured mesh coarsening algorithm has been developed for use in conjunction with multilevel methods. The algorithm preserves geometrical and topological features of the domain, and retains a maximal independent set of interior vertices to produce good coarse mesh quality. In anisotropic meshes, vertex selection is designed to retain the structure of the anisotropic mesh while reducing cell aspect ratio. Vertices are removed incrementally by contracting edges to zero length. Each vertex is removed by contracting the edge that maximizes the minimum sine of the dihedral angles of cells affected by the edge contraction. Rarely, a vertex slated for removal from the mesh cannot be removed; the success rate for vertex removal is typically 99.9% or more. For two‐dimensional meshes, both isotropic and anisotropic, the new approach is an unqualified success, removing all rejected vertices and producing output meshes of high quality; mesh quality degrades only when most vertices lie on the boundary. Three‐dimensional isotropic meshes are also coarsened successfully, provided that there is no difficulty distinguishing corners in the geometry from coarsely‐resolved curved surfaces; sophisticated discrete computational geometry techniques appear necessary to make that distinction. Three‐dimensional anisotropic cases are still problematic because of tight constraints on legal mesh connectivity. More work is required to either improve edge contraction choices or to develop an alternative strategy for mesh coarsening for three‐dimensional anisotropic meshes. Copyright © 2003 John Wiley & Sons, Ltd.