激光光条中心线提取研究综述

激光光条中心线提取在视觉测量、三维重建等领域具有重要的作用.介绍了不同类型的中心线提取模型,并且回顾了这些模型的转变和创新.具体来说,根据模型采用的核心算法,将中心线提取模型分为传统提取模型和基于深度学习的提取模型;传统中心线提取模型又分为极值模型、灰度重心模型、曲线拟合模型、基于Hessian矩阵的Steger模型和...     

动态环境下的激光条纹中心线提取方法

线结构光三维视觉测量技术最关键的一步是提取出结构光图像中的激光条纹中心线;针对动态测量环境下激光条纹图像存在复杂背景信息、激光光强分布不均、光带各部分宽度差别大、激光条纹断裂等问题,文章研究了一种适用于动态测量环境的激光条纹中心线提取方法;首先通过图像预处理以及自适应裁剪算法提取出感兴趣区域(ROI,region of interest);其次通过改进型伽马校正(IGC,improved gamma correction)以及改进型变阈值大津阈值分割算法(IVT Ostu,improved variable threshold)分割出激光条纹区域;然后使用二维灰度重心法(TD-GBM,two-dimensional gray baryeentric method)提取激光条纹的初始中心线;最终使用二次优化算法对初始中心线进行优化,精确地提取出激光条纹中心线;实验结果表明,相比于灰度重心法、Steger法等算法,文章所提方法受背景干扰以及激光条纹质量的影响较小,能够在多种复杂情况下更精确地提取激光条纹中心线,满足准确性高、稳定性强以及实时性好的要求.     

大曲率激光条纹中心线提取方法研究

激光三维扫描中线激光条纹中心位置的确定直接决定了检测的精度。检测系统需要在恶劣工况下快速完成图像采集和处理,精度和实时性要求高。在铆钉表面激光条纹曲率变化大且实时性要求高的情况下,采用了基于并行形态学细化和方向模板的灰度重心法,对图像光条ROI进行运算,准确高速提取激光条纹中心线。该方法检测速度快、精度高,满足检测系统设计要求。     

结肠中心线快速提取算法研究

中心线的提取速度是提高结肠计算机辅助检测的效率的重要因素.为此提出了一种基于生成树的中心线快速提取算法(FMST).在分析了最大生成树中心线提取算法(MST)存在大量冗余数据特点的基础上,利用边主源辅的搜索策略,保留趋于物体中心的关键点,通过消减冗余数据的方法来提高MST算法的速度.在10套已知中心线金标准的结肠仿真数据和2套结肠CT数据上的实验结果表明,FMST算法加快了MST算法的提取速度并且保持了中心线提取的准确性;在仿真数据上,FMST算法较MST算法的速度提高了80％以上,中心线重合率达到96.98％.     

CT冠脉造影中冠状动脉中心线的提取

提出了一种冠状动脉中心线的提取方法,该方法以FastMarching(快速行进)算法为基础,首先对图像进行下采样,接着使用血管增强滤波器对图像进行滤波,然后利用FastMarching算法提取下采样图像中初始点之间的最短路径,并使用最短路径在原始图像上计算冠状动脉中心线。在公共数据集上的验证实验表明,该方法具有较好的鲁棒性和准确性。     

基于最小代价路径的血管中心线提取

为解决传统最小代价路径算法提取血管中心线时存在偏向血管侧壁的问题,提出一种基于点的中心线校正方法。应用最小代价路径算法初步提取中心线,然后根据血管剖面灰阶值呈高斯分布的特点对每个中心点进行校正,再利用三次B样条将离散的中心点拟合为一条连续的中心线。实验结果表明,该算法提取的中心线更靠近血管的中心处,且对噪声具有鲁棒性。此外,将该算法用于起点、终点位置的校正,则提取的中心线对用户定义点的位置不敏感。     

一种基于脊线跟踪的冠状动脉中心线提取方法

冠脉血管中心线的提取是血管造影图像定量分析中的关键步骤。基于脊线跟踪法，提出了一种血管中心线自动提取方法。通过交互式地指定一个起始点和一个终止点，该算法能够自动获取两点间的血管中心线。实验结果表明了该方法的鲁棒性和可重复性。     

基于格网相交的遥感影像道路中心线提取方法

为从遥感影像中获得可用于建模的道路中心线矢量数据,首先利用成熟光谱反射(面向对象图像分类)方法,自动判读出道路范围,继而使用转化工具得出道路范围的矢量轮廓;通过添加辅助线与轮廓相交、提取线段中心点的方式得出道路中心线上的点集;最后设计工作流对点集进行排序处理,最终转化成可以用于规则建模的道路矢量数据.结果表明上述方法可...     

应用于造影图像的血管中心线全自动提取方法

针对传统血管中心线提取方法计算量大、需大量人工介入操作的问题,提出一种全自动的血管中心线跟踪提取方法.首先根据血管尺度设计不同的Frobenius范数对血管进行自适应增强,并基于血管灰度分布构造多尺度微分算子以建立判别函数,从而获得种子点的初始位置和跟踪方向;然后在初始方向局部弧长范围内优化检测新的脊点与跟踪方向,以确保在有较大误差干扰的情况下依然能够获得正确的脊点位置;最后利用血管拓扑结构特征检测方法剔除伪血管中心线.实验结果证明,该方法无需人工介入即可准确地在冠脉造影图像中提取出血管中心线、方向矢量等信息,可被用于临床心血管疾病的计算机辅助诊疗过程.     

高分辨率航空图像的道路中心线提取法

在高分辨率航空或卫星图像中提取道路特征具有越来越重要的实际意义。道路与背景通常存在灰度差异,论文首先求取高斯平滑图像的偏导数,进而确定图像中的脊点,然后根据连接准则连线,再构造判别规则删除道路虚警,最后利用Canny边缘图修正道路中心线。算法还对交点区域进行了有效处理。实验证明,该方法在一定程度上能够消除树、建筑物阴影和汽车对道路提取造成的影响,提取的道路中心线比较准确,能为道路网检测的后续处理提供重要的数据支持。     

Adaptive Shape-Sensitive Meshing of the Medial Axis

The medial axis transform provides an alternative representation of geometric shape that has many useful properties for analysis modeling. Applications include decomposition of general solids into subregions for mapped meshing, identification of slender regions for dimensional reduction and recognition of small features for suppression. To serve these purposes effectively, it is important to be able to mesh the medial axis so that its geometry is adequately approximated. This paper describes a general idea, which is based on equal distance criteria, for adaptive mesh refinement on the medial axis, assuming its topology has been defined. The completed set of theories and examples for 2D planar objects and 3D solid objects is presented. ID="A1" Correspondence and offprint requests to: C. Armstrong, The Queen's University of Belfast, Ashby Building, Stranmillis Road, Belfast, BT9 5AH. E-mail: c.armstrong@qub.ac.uk     

Approximating the Medial Axis from the Voronoi Diagram with a Convergence Guarantee

The medial axis of a surface in 3D is the closure of all points that have two or more closest points on the surface. It is an essential geometric structure in a number of applications involving 3D geometric shapes. Since exact computation of the medial axis is difficult in general, efforts continue to improve their approximations. Voronoi diagrams turn out to be useful for this approximation. Although it is known that Voronoi vertices for a sample of points from a curve in 2D approximate its medial axis, a similar result does not hold in 3D. Recently, it has been discovered that only a subset of Voronoi vertices converge to the medial axis as sample density approaches infinity. However, most applications need a nondiscrete approximation as opposed to a discrete one. To date no known algorithm can compute this approximation straight from the Voronoi diagram with a guarantee of convergence. We present such an algorithm and its convergence analysis in this paper. One salient feature of the algorithm is that it is scale and density independent. Experimental results corroborate our theoretical claims.     

Medial axis of a planar region by offset self-intersections

The medial axis (MA) of a planar region is the locus of those maximal disks contained within its boundary. This entity has many CAD/CAM applications. Approximations based on the Voronoi diagram are efficient for linear-arc boundaries, but such constructions are more difficult if the boundary is free. This paper proposes an algorithm for free-form boundaries that uses the relation between MA and offsets. It takes the curvature information from the boundary in order to find the self-intersections of successive offset curves. These self-intersection points belong to the MA and can be interpolated to obtain an approximation in Bézier form. This method also approximates the medial axis transform by using the offset distance to each self-intersection.     

Medial Axis Construction in Three Dimensions and its Application to Mesh Generation

An algorithm for the construction of the medial axis of a three-dimensional body given by a triangulation of its bounding surface is described. The indirect construction is based on the Delaunay-triangulation of a set of sample points on the bounding surface. The point set is refined automatically so as to capture the correct topology of the medial axis. The computed medial axis (or better medial surface) is then used for hex-dominant mesh generation. Quad-dominant meshes are generated on the medial subfaces first and extruded to the boundary of the body at both sides. The resulting single cell layer is subdivided in direction normal to the boundary, yielding columns of hexahedral and three-sided prismatic cells. The resulting volume mesh is orthogonal at the boundary and ‘semi-structured’ between boundary and medial surface. Mixed cell types (tets, pyramids, degenerate hexahedra) may result along the medial surface. An advancing front algorithm (paving) is used for meshing the subfaces of the medial surface. Development of the mesh generator has not been fully completed with respect to degenerate parts of the medial axis. First medium-complexity bodies have been meshed, however, showing moderate meshing times.     

A Recursive Digital Filter Implementation for Noisy and Blurred Images

This paper presents the design and the implementation of an optimized Canny-Deriche edge detector. After a brief reminder of the filter's equations, we define different techniques to speed up the sampling rate of the IIR filter. In particular, improving the throughput rate of the IIR filter, we present a look-ahead with decomposition technique. This method leads us to design a first chip, which performs at a sampling rate of over 20 MHz with a silicon area of 60 mm². Using a local register retiming method, we have designed a second circuit, which is able to process a pixel in 30 ns with a silicon area of 30 mm². These two approaches are compared. This work leads us to an ASIC which was designed in a CMOS 1 μm technology and successfully tested.     

Area-Based Medial Axis of Planar Curves

A new definition of affine invariant medial axis of planar closed curves is introduced. A point belongs to the affine medial axis if and only if it is equidistant from at least two points of the curve, with the distance being a minimum and given by the areas between the curve and its corresponding chords. The medial axis is robust, eliminating the need for curve denoising. In a dynamical interpretation of this affine medial axis, the medial axis points are the affine shock positions of the affine erosion of the curve. We propose a simple method to compute the medial axis and give examples. We also demonstrate how to use this method to detect affine skew symmetry in real images.     

多边形Voronoi图和中轴的特征从属性证明

VD偏重于对图形的精确描述,MA则偏重于对图形整体的简洁描述。为适用更广泛的工程应用,分析对比了含直线段、圆弧和自由曲线的区域轮廓边界Voronoi图和中轴的各自的特征,给出了Voronoi图与中轴之间不同情况下特征从属性的证明。     

实体模型的多分辨率中轴生成

针对模型降维的需要,提出一种实体模型的多分辨率中轴生成方法.首先根据模型的层次寻找模型中需要细化的部分,并据此对边界进行体素化;其次使用距离蔓延的方法计算多分辨率体素的距离;最后根据体素的距离计算出多分辨率的中轴体素.实验结果表明,该方法可以以低时间空间代价生成高质量的多分辨率中轴.     

基于两幅模糊与噪声图像的图像交替修复算法

针对一幅模糊图像和一幅噪声图像的图像修复问题,提出了一种结合TV-L~1模型与TV模型的新交替修复算法.该算法首先利用TV-L~1模型对噪声图像进行去噪;然后利用TV模型,把去噪结果作为迭代初始值,对于模糊图像去模糊;最后,把去模糊结果作为迭代初始值,再利用TV-L~1模型对噪声图像进行去噪……,如此交替进行.实验效果表明,该新算法不仅继承了TV-L~1模型与TV模型能保持轮廓和细节的优点,同时也有效地克服了这两种模型会降低对比度和出现"重影"的缺点.     

二维Voronoi图和中轴的特征区分

对多边形Voronoi图和中轴两个概念进行了重新定义以适应更广泛的工程应用.分析对比了含直线段、圆弧和自由曲线的区域轮廓边界Voronoi图和中轴的各自的特征,给出了二者在凸多边形、简单多边形等不同情况下的相互关系及证明.