基于类似形原理匹配轴测图

依据轴测图中存在的仿射不变性 ,提出 5个描述多边形拓扑和几何特征的特征向量 ,以这些特征向量为约束条件给出轴测投影下类似形的定义。提出基于类似形原理匹配轴测图新方法。基于类似形原理可将轴测图中的面分类。匹配过程限制在类似形基础上 ,可大大减少搜索范围。用拓扑结构定性分析类似形对图像中的噪声不敏感 ,用几何特征进行定量分析能反应轴测图的细小差别。该匹配算法能识别不同形状的目标 ,而且还可以区分形状相同但大小、比例不同的目标。实验结果表明该匹配方法是有效的。     

一种基于分裂合并的多边形逼近算法

断层医学图象目标组织经图象分割、轮廓跟踪后得到的轮廓像素点数据量较大,不宜直接用于几何建模。多边形逼近是提取曲线特征点和简化数据、加快图形运算的一个重要方法,提出一种基于分裂合并的多边形逼近算法,将轮廓像素点集合分段进行线段逼近,逐次递增进行共线检查,反复执行分裂、合并操作,直到所有逼近误差在指定范围内,逼近多边形不再改变为止。该算法能够在保持原始轮廓形状特征的前提下,有效减少数据量,提高了计算效率。     

推镗深孔产生多边形的原因及对策

一、概述以电站转子的中心孔为例，其内孔不允许多边形存在，否则，转速为3000r／min的转子所产生的离心力，能从多边形这个薄弱环节造成转子损坏。因此，研究多边形产生的原因及对策有重大意义。螺旋状多边几何形见图1，可以看出多边形边缘上任意一点如凹进去，则对面的一点就凸出来，且凹和凸点必然通过多边形孔的中心。可见多边形的几何形状是以中心对称。用内径百分表测量孔径，孔径任何方向都相等。因此，用检查孔径的方法很难发现是多边形孔。图1多边形几何形状（横截面）a．3边形b．5边形c．7边形多边形孔在推镗切削过程中逐渐形成，…     

手指宽度选取与识别率对应关系分析

手指宽度是手形识别的有效特征之一,而选取的测量点个数影响识别的准确率和时间.本文研究了选取手指宽度的个数与识别率、匹配时间的对应关系,以用于根据不同的识别率要求估计不同人手选取的宽度个数.实验论证了在定位准确、每个手指选取的宽度个数不少于3个的条件下,识别率可以达到90%以上.提出了自动识别算法,利用曲率法定位特征点,根据特征点计算手指的绝对宽度,应用绝对宽度之间的相对宽度构成特征向量,计算特征向量之间的欧式距离,实现身份识别.实验获得了手指宽度个数与匹配时间的对应关系以及手指宽度个数与识别率的对应关系曲线.结果表明,每个手指选取6个宽度时,就可以获得区分个体的特征量,识别率达到94.03%.     

看图方法研究--类似形互补原理

提出了一种利用类似形互补原理，判断两个不同投影面的垂直面是否相交以及它们的交线部分的看图新方法，目的在于解决视图中的斜线问题，便于理解和把握形体的整体形状，便于看图和画图。尤其是当形体比较复杂，甚至经多次切割时，更能体现出此方法的妙用。     

三维目标离面旋转和非线性光照识别研究

针对三维目标在离面旋转以及非线性光照条件下的识别问题,采集了180幅不同旋转角度的图像作为训练集,利用主分量分析法选取了20个特征向量,显著降低了特征空间的维数.利用这些特征向量对图像进行分解和重构,在保持较高计算精度的前提下减少了计算的复杂性.提出了利用原始图像向量与重构图像向量夹角余弦值来判别目标真假的方法.模拟结果表明,该方法能有效识别不同旋转角度的目标,同时能够消除非线性光照对目标识别的影响.     

装配体ICT图象的内外边界识别研究

用小波变换提取装配体ＩＣＴ图象中的边界轮廓,提出基于简单环识别装配体ＩＣＴ图象的内外边界,该算法可用于装配体的结构形状分析并能正确地建立实体模型。     

基于迭代修正DFT的车轮多边形磨耗状态识别

受车辆变速运行、轨道随机不平顺等因素的影响,轨道车辆的运行过程为典型的非平稳过程。由于传统方法对于非平稳信号的处理不理想,使得车轮多边形磨耗阶次与幅值的准确识别比较困难。为提高车轮多边形磨耗状态的识别准确度,提出了一种基于迭代修正离散傅里叶变换（discrete Fourier transform,简称DFT）的动态识别方法,采用车辆处于相对稳定速度运行时产生的轴箱垂向振动加速度信号进行分析。首先,通过设定适当的平稳性检验条件,从样本信号中抽取出部分相对平稳的短时信号;其次,对所抽取的短时信号进行频域分析与迭代计算,获得各阶车轮多边形的振动频率与振动周期;然后,根据振动周期的长度对所抽取的短时信号进行二次截断,获得代表各阶车轮多变形振动周期整数倍长度的新短时信号;最后,结合车轮多边形的几何特征与动态特性,对抽取的新短时信号再次进行频域分析与磨耗参数（阶次、幅值）计算,进而实现对车轮多边形磨耗状态的准确识别。分析表明,该识别方法有效减少了传统分析方法中因栅栏效应和频谱泄漏等固有缺陷导致的识别误差,可消除大部分非平稳因素的干扰,为轨道车辆的安全运行维护提供理论支持和方法参考。     

基于正交特征的手写体汉字识别方法

手写体汉字识别一直是模式识别研究领域中的难点.本文将矩变换与弹性网格方法结合起来,用正交Zernike矩提取全局特征,用弹性网格提取局部特征,然后用K-L变换对局部特征向量正交化.这样得到的特征向量不仅具有汉字的全局和局部特征,而且是最有分辨能力的特征向量.实验证明该方法是行之有效的.     

一种基于视觉形状模型的机器人目标识别算法

为实现机器人主动识别并抓取预定目标,提出了一种基于形状模型的目标识别算法。以轮廓质心为基准,极半径生成目标的形状特征向量,然后以高斯模型完成样本特征模型库的训练。采用动态时间规整DTW(dynamic time warping)解决特征向量元间的匹配问题,并依据惯量力矩(moment of inertia)获取形状的主轴参数,给出了特征向量起始次序的对正方法。实验表明该算法可以在较复杂环境下识别预定目标,不受平移、尺度与旋转几何变化的影响,具有较高的鲁棒性。     

体素几何模型中物体表面的绘制算法

在体素几何模型中 ,物体由体素表示 ,物体表面由边界体素表示。当体素大小与屏幕像素大小相等时 ,可把边界体素作为点向投影面投影 ,但这样易产生空洞现象和图像走样。基于物体边界体素 ,生成一个绘制表面 ,用光线投射法显示这个绘制表面 ,可以消除空洞现象 ,减轻图像走样。这种算法能生成光滑细腻的图像。本文生成绘制表面的速度比 Dividing Cubes算法生成表面的速度快     

微型精密零件图像测量清晰度算法的选择与综合评价算法

随着仪器、设备的日益小型化,微型精密零件应用日益广泛。由于常规测量手段的局限性,光学影像测量方法在微型精密零件测量中成为国内外竞相发展的技术。然而在图像采集过程中对图像清晰度的评价算法虽种类较多但各算法之间缺乏统一。本论文基于微型精密零件的成像测量方法,使用前人制作的图像测量仪,使用远心成像测量系统在不同工作距离处进行图像采集,然后使用相关算法进行清晰度评价,最后对所得数据进行主成分分析和因子分析,从8种图像清晰度评价算法计算出适用于课题使用的评价算法,并建立了综合清晰度评价模型。本文的数据处理模式能够对算法进行选择并解决不同算法间的协调问题。而且本文提出的数据挖掘算法能针对同一要求的不同评价值进行选择及综合,同时能大幅减少后续运算量。     

A neural network approach to the classification of 3D prismatic parts

This paper presents a neural network approach to the classification of 3D prismatic parts based on their global shape information modelling. In this approach, a 3D part is modelled by the contours of its three projected views, which are approximately represented by three rectilinear polygons. The global shape information of each polygon is modelled by its simplified skeleton, which originally is of a tree structure and can be represented by several vectors by a conversion method. These vectors are the input to a polygon classifier which is constructed on the basis of the back-propagation neural network model. The classification results of polygons can be used to group the 3D prismatic parts into families in a hierarchical manner, by setting different levels of similarity criteria. The proposed method for classifying 3D workpieces can be used to enhance the productivity of design and manufacturing processes. By retrieving and reviewing similar parts from the part families, the designers or process planners could be greatly assisted in performing a new task. That is, they can avoid the reinvention of an existing design and can create a new design by modifying existing ones.     

Area Variation Based Color Snake Algorithm for Moving Object Tracking

A snake algorithm has been known that it has a strong point in extracting the exact contour of an object.But it is apt to be influenced by scattered edges around the control points.Since the shape of a moving object in 2D image changes a lot due ta its rotation and translation in the 3D space,the conventional algorithm that takes into account slowly moving objects cannot provide an appropriate solution.To utilize the advantages of the snake algrithm while minimizing the drawbacks,this paper proposes the area variation based color snake algorithm for moving object tracking.The proposed algorithm includes a new energy term which is used for preserving the shape of an object between two consecutive inages.The proposed one can also segment precisely interesting objects on complex image since it is based on color information.Experiment results show that the proposed algorithm is very effective in various environments.     

MRT letter: Recovering weak‐textured surfaces using image focus

In nature, objects have partially weak texture and their shape reconstruction using focus based passive methods like shape from focus (SFF), is difficult. This article presents a new SFF algorithm which can compute precise depth of dense as well as weak textured objects. Segmentation is applied to discard wrong depth estimate and then later interpolating them from accurate depth values of their neighbors. The performance of the proposed method is tested, using different image sequences of synthetic and real objects, with varying textures. Microsc. Res. Tech., 2009. © 2009 Wiley‐Liss, Inc.     

废有色金属分选中散布物料动态位置的实时计算

针对废有色金属分选中物料动态位置计算的问题,提出了一种基于二值图像形心计算的改进算法,以保证物料位置计算的准确性和高效性。对基于X射线透射技术的废有色金属分选进行了分析,利用图像处理技术对探测数据进行了处理。考虑到待分选有色金属废料形状、尺寸、姿态各异,移动速度快的特点,选用了高精度的连通区标记法,并在二值图像形心算法的基础上进行了改进,实现了连通区标记后一次遍历计算多个物料目标形心,解决了同一物料的探测数据被分割至两幅图的问题,完成了物料位置计算。实验结果表明,该计算方法能快速、准确地计算有色金属物料的动态位置,算法的准确率可达92.3%,平均每幅图耗时30.3 ms。     

Local Mesh Decimation for View-Independent Three-Axis NC Milling Simulation

The view dependency of typical z-buffer-based NC simulation methods is overcome by a polygon rendering technique that generates polygons to represent the workpiece, thus enabling dynamic viewing transformations without reconstruction of the entire data structure. However, the polygon rendering technique still has difficulty in realising real-time simulation owing to the limited performance of current graphics devices. There-fore, it is necessary to develop a mesh decimation method that enables rapid rendering without loss of display quality. To meet this requirement, in this paper, a new mesh decimation algorithm is proposed for a workpiece which is represented by the z-map structure and whose shape is dynamically changed by the tool. In this algorithm, the z-map data for the workpiece are divided into several regions, and a triangular mesh is constructed for each region. Then, if the tool cuts any region, its mesh is regenerated and decimated again. Since the range of mesh decimation is confined to a few regions, the reduced triangles for rapid rendering can be obtained in a short time. This local mesh decimation method may contribute toward realising the polygon rendering-based NC simulation in real-time on computers equipped with general graphics cards.     

Geometrical characterization of fluorescently labelled surfaces from noisy 3D microscopy data

Modern fluorescence microscopy enables fast 3D imaging of biological and inert systems alike. In many studies, it is important to detect the surface of objects and quantitatively characterize its local geometry, including its mean curvature. We present a fully automated algorithm to determine the location and curvatures of an object from 3D fluorescence images, such as those obtained using confocal or light‐sheet microscopy. The algorithm aims at reconstructing surface labelled objects with spherical topology and mild deformations from the spherical geometry with high accuracy, rather than reconstructing arbitrarily deformed objects with lower fidelity. Using both synthetic data with known geometrical characteristics and experimental data of spherical objects, we characterize the algorithm's accuracy over the range of conditions and parameters typically encountered in 3D fluorescence imaging. We show that the algorithm can detect the location of the surface and obtain a map of local mean curvatures with relative errors typically below 2% and 20%, respectively, even in the presence of substantial levels of noise. Finally, we apply this algorithm to analyse the shape and curvature map of fluorescently labelled oil droplets embedded within multicellular aggregates and deformed by cellular forces.