基于轮廓的多设备医学图象的刚性配准

提出了一种基于轮廓的多设备医学图象的刚性配准算法。在半自动轮廓搜索的基础上，运用迭代最近邻点方法搜索轮廓间的对应点，最小化代价函数并获得最佳刚性变换，实现图象的精确配准。通过多次选择初始变换的搜索策略，解决了搜索过程中出现的局部极小值问题。实验表明该方法能有效地实现多设备医学图象的配准。     

基于联合直方图区域计数的时间序列图象自动配准方法

时间序列图象的配准是医学临床科研、诊断和治疗中图象应用的必要步骤。为了快速、准确、简便地进行医学序列图象配准，提出了一种新的用于时间序列图象自动配准的方法，该方法利用了图象的联合直方图，首先通过对图象做简单的阈值分割，将联合直方图划分为4个区域，然后根据不同的配准图象数据选择定义在不同区域上的计数值作为参数计算的准则函数。该方法设计简单、巧妙，以计数方法代替其他方法中大量的浮点运算。由于准则函数具有良好的光滑特性，且选择Powell算法做最优化搜索，因此保证了优化结果的准确性。和其他算法相经，该方法大大简化了准则函数的计算，从而显著提高了配准优化搜索的速度。根据实验结果及基于互信息量方法做的对比，证明本文提出的方法准确、简便、快速、有效。     

模糊多准则图象重建技术

本文提出一种适合频谱数据成象的多准则优化成象算法，由于数字图象品质评价的模糊性，所以，我们从模糊数学理论出发，提出一套模糊准则函数，这些准则函数多准则优化可以在频域进行，它可用于核磁共振图象重建，实际和模拟结果都表明重建图象质量高于现有方法。     

煤矿多媒体安全监视系统中基于IFS的图象压缩方法探讨

概述了分形及分形图象压缩编码的理论基础，讨论了基于迭代函数系统和新的分形图象压缩编码。     

动态步长FOA的PID参数整定方法研究

为了解决基本果蝇优化算法(FOA)因固定搜索步长而对比例积分微分(PID)参数整定收敛精度不高且搜寻效率低的问题,将Logistic(t)的变换函数lgt(t)引入FOA中。由该变换函数确定自适应步长,提出一种动态步长果蝇优化算法(DSFOA)。DSFOA中果蝇个体搜索步长会随着迭代次数的增加而动态地变化。该算法在迭代前期使用大步长,具有更高的全局搜索效率;在迭代后期使用小步长,具有较强的局部寻优能力。这可以提高收敛精度,实现对全局搜索和局部搜索过程的优化。二阶系统仿真测试结果表明,相比于FOA,DSFOA寻优过程产生的PID参数使系统性能更优,能快速、有效地搜索到PID最优参数且鲁棒性好。该结果验证了DSFOA的有效性与合理性。     

基于迭代分形的图象压缩和检索方法

图象所具有的海量性和无序性的特点，决定了多媒体应用的构建必须解决图象数据的高效压缩和有效检索两个关键问题，而由于传统的压缩和检索技术的研究是相互分离的，因而限制了多媒体应用系统整体性能的提高，针对此问题，从两者相互结合的观点，对图象压缩和检索方法进行了研究，首先在小波变换域内，基于迭代分形对图象数据进行压缩，然后在图象分形码的基础上，利用迭代函数系统分布特性构建的特征量来支持图象检索，实验结果验证了该方法的可行性和有效性，同时也表明了基于迭代分形的图象检索方法所具有的巨大应用潜力。     

手写字符串识别搜索算法

字符串识别通过最优路径搜索得到字符切分和字符识别结果.本文将字符同步和时间同步两种搜索模式应用于手写字符串识别系统,比较两种模式下使用不同准则函数和搜索算法的系统性能.同时,提出一种改进的路径评价准则,在此准则下可用动态规划算法进行最优路径的搜索.在联机手写日文字符串识别中的实验结果表明.对于无词典驱动的字符串识别系统,时间同步搜索的效率高于字符同步搜索.利用本文所提出的路径评价准则,可得到与归一化准则相当的切分和识别准确率,但搜索时间大为减少.     

迭代步进值自适应调整的果蝇优化算法

针对传统果蝇优化算法（FOA）收敛精度不高和易陷入局部最优的缺点,提出了一种迭代步进值自适应调整的果蝇优化算法（FOAMR）。在该算法中,引入了果蝇群体速度进化因子和聚集度因子,并将迭代步进值表示为以上2个参数的函数同时定义自适应调整因子。在每次迭代时,算法根据当前果蝇群体速度进化因子和聚集度因子动态调整步进值的大小并通过自适应调整因子动态调整搜索距离的大小。对典型函数的测试结果表明,FOAMR比FOA具有更好的全局搜索能力,同时收敛速度、收敛精度明显提高。     

基于IFS理论的快速图象编码和解码算法

分形图象编码法具有高压缩比、高图象品质等显著优点，但其编码时间太长。文中提出的快速编码算法，用偏差图导出距离度量，用二叉树和链表结构进行搜索匹配，可使编码速度提高几十倍乃至百倍。同时，用偏差图进行解码迭代，可使解码过程加速，而还原图象质量不变。     

求解高维函数优化的动态粒子群算法

针对基本粒子群优化算法对高维函数优化时搜索精度不高的缺陷,提出了一种动态粒子群优化算法。该算法采用了通过调节阈值对粒子运动轨迹进行动态改变的策略,使得粒子对周围环境的适应能力不受进化代数的影响,从而保证了算法在迭代后期仍具有较强的搜索能力。实验结果表明,与文献算法相比,该算法在处理高维函数优化时具有更强的寻优能力和更高的搜索精度。     

边界简化与多目标优化相结合的高质量四边形网格生成

目的 高质量四边形网格生成是计算机辅助设计、等几何分析与图形学领域中一个富有挑战性的重要问题。针对这一问题,提出一种基于边界简化与多目标优化的高质量四边形网格生成新框架。方法 首先针对亏格非零的平面区域,提出一种将多连通区域转化为单连通区域的方法,可生成高质量的插入边界;其次,提出"可简化角度"和"可简化面积比率"两个阈值概念,从顶点夹角和顶点三角形面积入手,将给定的多边形边界简化为粗糙多边形;然后对边界简化得到的粗糙多边形进行子域分解,并确定每个子域内的网格顶点连接信息;最后提出四边形网格的均匀性和正交性度量目标函数,并通过多目标非线性优化技术确定网格内部顶点的几何位置。结果 在同样的离散边界下,本文方法与现有方法所生成的四边网格相比,所生成的四边网格顶点和单元总数目较少,网格单元质量基本类似,计算时间成本大致相同,但奇异点数目可减少70% 80%,衡量网格单元质量的比例雅克比值等相关指标均有所提高。结论 本文所提出的四边形网格生成方法能够有效减少网格中的奇异点数目,并可生成具有良好光滑性、均匀性和正交性的高质量四边形网格,非常适用于工程分析和动画仿真。     

Automatic generation of finite element meshes

This paper describes a new algorithm for the automatic generation of finite element meshes of arbitrary multiply connected domains. The strategy is based upon the construction of a mapping from the generated mesh into a regular one. The scheme is designed for maximum flexibility and is able to generate meshes of triangular or quadrilateral curved elements. Several examples are presented to illustrate the applicability of the algorithm.     

Generating Symbolic Interpolants for Scattered Data with Normal Vectors

Algorithms to generate a triangular or a quadrilateral interpolant with G1-continuity are given in this paper for arbitrary scattered data with associated normal vectors over a prescribed triangular or quadrilateral decomposition. The interpolants are constructed with a general method to generate surfaces from moving Bezier curves under geometric constraints. With the algorithm, we may obtain interpolants in complete symbolic parametric forms, leading to a fast computation of the interpolant. A dynamic interpolation solid modelling software package DISM is implemented based on the algorithm which can be used to generate and manipulate solid objects in an interactive way.     

Generation of bi-monotone patches from quadrilateral mesh for reverse engineering

Thanks to recent improvements, computational methods can now be used to convert triangular meshes into quadrilateral meshes so that the quadrilateral elements capture well the principal curvature directional fields of surfaces and intrinsically have surface parametric values. In this study, a quadrilateral mesh generated using the mixed integer quadrangulation technique of Bommes et al. is used for input. We first segment a quadrilateral mesh into four-sided patches. The feature curves inside these patches are then detected and are constrained to act as the patch boundaries. Finally, the patch configuration is improved to generate large patches. The proposed method produces bi-monotone patches, which are appropriate for use in reverse engineering to capture the surface details of an object. A shape control parameter that can be adjusted by the user during the patch generation process is also provided to support the creation of patches with good bi-monotone shapes. This study mainly targets shape models of mechanical parts consisting of major smooth surfaces with feature curves between them.     

四边形网格间接生成方法

研究了基于背景三角网格的四边形网格间接生成算法,并针对三角形合并过程中容易残留三角形的缺陷提出了确定侧边的详细算法,该算法主要是依据背景三角网格中边的位置和前沿边的情形,通过背景三角网格中已存在的边、边交换或边分割确定侧边,以避免在三角形合并过程中残留三角形单元。最后给出实例验证了算法的有效性。     

A methodology for quadrilateral finite element mesh coarsening

High fidelity finite element modeling of continuum mechanics problems often requires using all quadrilateral or all hexahedral meshes. The efficiency of such models is often dependent upon the ability to adapt a mesh to the physics of the phenomena. Adapting a mesh requires the ability to both refine and/or coarsen the mesh. The algorithms available to refine and coarsen triangular and tetrahedral meshes are very robust and efficient. However, the ability to locally and conformally refine or coarsen all quadrilateral and all hexahedral meshes presents many difficulties. Some research has been done on localized conformal refinement of quadrilateral and hexahedral meshes. However, little work has been done on localized conformal coarsening of quadrilateral and hexahedral meshes. A general method which provides both localized conformal coarsening and refinement for quadrilateral meshes is presented in this paper. This method is based on restructuring the mesh with simplex manipulations to the dual of the mesh. In addition, this method appears to be extensible to hexahedral meshes in three dimensions. Sandia National Laboratories is a multiprogram laboratory operated by Sandia Corporation, a Lockheed Martin Company, for the United States Department of Energy under Contract DE-AC04-94AL85000.     

Medial surface generation using chordal axis transformation in shell structures

This paper describes the generation of chordal surfaces for various shell structures, such as automobile bodies, plastic injection mold components, and sheet metal parts. After a single-layered tetrahedral mesh is generated by the advancing front method, the chordal surface is generated by cutting the tetrahedral mesh. One or more shell elements are generated at each tetrahedral element, and the chordal surface is constructed with triangular or quadrilateral shell elements. This algorithm has been tested on several models with rib structures.     

A two-dimensional paving mesh generator for triangles with controllable aspect ratio and quadrilaterals with high quality

Certain classes of problems result in solution fields of which the characteristic length scales vary with the orientation. Often the orientation of these length scales is related to the orientation of the boundaries. Such solution fields can be captured by the finite element method, using a mesh that is refined in the direction of the short length scales and coarse in the other directions. These meshes contain elements with high aspect ratios in a predefined pattern.The mesh generator presented here can render triangles with high aspect ratios through a paving algorithm. The paving algorithm that is employed applies both triangles and quadrilaterals, combining the advantages of both to render a qualitatively good, oriented triangular mesh, with a concentration of elements in the direction where the internal length scales of the solution field are the shortest.The mesh generator produces triangles with one (almost) orthogonal corner. When low aspect ratio triangles are generated, these are well suited for conversion to quadrilateral elements. Test results indicate that quadrilateral meshes converted from the mesh generator introduced here have a considerably better quality than those converted from several other triangular mesh generators.     

An algorithm for automatic 2D quadrilateral mesh generation with line constraints

Finite element method (FEM) is a fundamental numerical analysis technique widely used in engineering applications. Although state-of-the-art hardware has reduced the solving time, which accounts for a small portion of the overall FEM analysis time, the relative time needed to build mesh models has been increasing. In particular, mesh models that must model stiffeners, those features that are attached to the plate in a ship structure, are imposed with line constraints and other constraints such as holes. To automatically generate a 2D quadrilateral mesh with the line constraints, an extended algorithm to handle line constraints is proposed based on the constrained Delaunay triangulation and Q-Morph algorithm. The performance of the proposed algorithm is evaluated, and numerical results of our proposed algorithm are presented.     

Interpolation by geometric algorithm

We present a novel geometric algorithm to construct a smooth surface that interpolates a triangular or a quadrilateral mesh of arbitrary topological type formed by n vertices. Although our method can be applied to B-spline surfaces and subdivision surfaces of all kinds, we illustrate our algorithm focusing on Loop subdivision surfaces as most of the meshes are in triangular form. We start our algorithm by assuming that the given triangular mesh is a control net of a Loop subdivision surface. The control points are iteratively updated globally by a simple local point-surface distance computation and an offsetting procedure without solving a linear system. The complexity of our algorithm is O(mn) where n is the number of vertices and m is the number of iterations. The number of iterations m depends on the fineness of the mesh and accuracy required.