利用加权有向图的船舶水尺重建与水位识别

针对目前船舶水位识别精度低、检测场景迁移性差等问题,提出一种基于加权有向图的高精度、鲁棒性好的船舶水尺重建和水位识别方法.首先通过目标检测和图像分割算法识别船体上的水尺字符和水位线;然后根据船舶水尺的结构特点将识别出来的水尺字符构建加权有向图,通过寻找加权有向图中的最长路径完成船舶的水尺重建;最后根据重建水尺和水位线的位置关系识别船舶的水位高度.在黄骅港实地采集的视频数据上进行实验的结果表明,所提方法可以较好地修正前期基于水尺字符检测网络带来的字符误检问题,大大减少因船舶图像背景复杂产生的不利影响,在10 mm的误差范围内,水位识别准确率可达91.3%,显著优于业内主流方法.     

民用航空飞机注册号自动识别

为了提高机场自动化水平,提出一种简单、有效的飞机注册号自动识别算法.对于机场摄像机获取的彩色跑道图像,首先采用Canny操作和轮廓检测提取飞机图像,为了便于计算机处理将其转换为灰度图像;然后使用Hough转换和SWT(Stroke Width Transform)文本定位法定位飞机尾部的注册号;最后应用阈值法分割注册号,使用基于孔洞的字符模板匹配法识别注册号字符.实验结果表明,该算法具有良好的飞机注册号识别效果,较强的鲁棒性,具有广泛的应用前景.     

基于组合特征的集装箱箱号识别法

为准确、高效地识别集装箱箱号,提出一种基于组合特征实现箱号识别的方法。对分割好的字符二值图像预处理后提取孔洞特征、凹凸特征、跳变特征、笔画特征等,利用树形分类器进行分类识别,并结合箱号自身的校验规则进行验证、识别。该方法不需要对字符图像做复杂的细化处理,提高了运算速度,避免了细化造成的字符畸变。实验结果表明,该方法平均箱号正确识别率可以达到93%,每箱号(11个字符)平均识别时间为0.065s。     

融合特征和先验知识的车牌字符图像检测算法

提出一种融合车牌字符切割后的二值字符图像的结构特征及对应的彩色小字符图像的颜色信息对车牌分割后的小字符图像进行真伪字符图像区分,以此达到检测字符图像目的的算法。为满足实时车牌对时间响应的要求,对车牌字符切割得到的灰度图像作快速二值化,在二值化的图像上提取结构信息,结合车牌字符分布的特点去除了大部分非颜色车牌的伪字符图像。对于难以从结构上进行字符检测的颜色车牌伪字符图像,在结构特征分析的基础上再次通过从彩色图像提取的颜色信息进行相似性分析,排除伪字符图像。对候选字符图像融合大间隔这个先验知识得到输出的字符图像。实验结果表明算法有良好的字符检测效果,可以用于实时车牌识别系统中作为字符切割后期处理一部分。     

印刷哈萨克斯拉夫粘连字符的切分方法

哈萨克斯拉夫图像文本经过行切分和列切分后,存在水平方向接触和垂直方向重叠的粘连字符。为提高字符识别率,依据字符连通域的最小外接矩形切分开垂直方向重叠的粘连字符图像块;利用判决条件:字符宽度概率密度分布图、字符图像块垂直投影的波峰数目和字符图像块垂直投影波峰的对称性,分离初始粘连字符图像块中正确的单个字符图像块和实际接触的粘连字符图像块;在允许的字符宽度范围内,寻找粘连字符图像垂直投影图的极小值点,以切分实际接触的粘连字符。实验结果表明,该方法泛化能力较好且识别率有明显提高。     

基于KAP有向图模型的医学图像分类算法*

脑部CT图像拥有良好的纹理特性且图像间纹理角点的位置近似相同。基于此原因,文中提出基于K最近邻纹理角点(KAP)有向图模型的医学图像分类算法。首先提出面向纹理的角点提取方法;然后针对提取的角点,结合医学图像的固有特点,提出KAP有向图模型用于描述医学图像;最后基于KAP有向图模型提出医学图像分类算法。实验表明,文中算法在时间复杂度和准确度方面都取得较好结果。     

一种改进的模糊连接度图像分割方法

本文对模糊连接度的理论框架进行了扩展,在框架内引入了图论中的有向图和森林等概念。把图像定义为一个有向图,有向图的结点是图像的像素点,有向图的弧是相邻像素的边,树的根结点是人工给出的种子点,所有像素点到种子点的模糊连接路径组成了模糊连接森林,模糊连接森林给出了图像的分割。实验结果表明本文方法比相对模糊连接度方法得到更好的结果。     

工件打码图像中的字符倾斜矫正算法研究

本文主要在模式识别领域,研究人工打码字符图像在图像预处理之后,识别之前的单个整体字符的倾斜矫正算法.对于实际的字符图像源,由于是人工逐个字符进行打码,难免会出现字符倾斜的状况,再加上工件背景比较复杂,拍摄光照不均等原因,使得对字符图像的处理和倾斜矫正愈加困难.本文结合实际应用,在对字符图像进行前期的预处理和字符分割之后,再对分割后的单个字符进行倾斜矫正,实验取得了良好的效果,为字符后期的识别创造了有利的条件.     

基于图像数据挖掘的有向图模型检索方法

为充分挖掘图像数据信息,提出了一种有向图模型检索方法,结合距离测度初次检索和有向图距离二次检索提高图像检索性能。首先,采用传统的纹理、边缘和颜色特征以及特征之间的欧氏距离测度来进行初次检索,得到一个查询排序列表;在此基础上,结合距离测度与余弦测度设计图像之间的相关测度,在不同的相关测度阈值下构建图像数据集的有向图模型集合;最后,计算有向图距离,据此进行二次检索,降低误检现象。在COREL和ImageCLEF两个数据集上的图像检索实验结果表明,该方法的平均精确度和平均召回率指标高。     

一种新的粘连字符图像分割方法

针对监控画面采样图像中数字的自动识别问题,提出一种新的粘连字符图像分割方法。该方法以预处理后二值图像的连通状况来判定字符粘连的存在,并对粘连字符图像采用上下轮廓极值法确定候选粘连分割点,以双向最短路径确定合适的图像分割线路。仿真实验表明,该方法能有效解决粘连字符图像的分割问题。     

The total domination and total bondage numbers of extended de Bruijn and Kautz digraphs

In this paper we consider the total domination number and the total bondage number for digraphs. The total bondage number, defined as the minimum number of edges whose removal enlarges the total domination number, measures to some extent the robustness of a network where a minimum total dominating set is required. We determine the total domination number and total bondage number of the extended de Burijn digraph and the extended Kautz digraph, proposed by Shibata and Gonda in 1995, which generalize the classical de Bruijn digraph and the Kautz digraph.     

基于有向图链式存储的关联规则改进算法

在关联规则数据挖掘领域中,Apriori算法是这个方面的经典算法,但它仍存在许多弊端,为此在Apriori算法的基础上提出了一种基于有向图链式存储的改进算法,此算法根据数据结构中有向图链式存储的结构,将所有事务全部存入链表,无需多次扫描数据库,只在事务链表中完成候选集和频繁集的寻找工作.此方法能够迅速得到候选集的支持度...     

逆向工程中复杂拓扑关系的三角网格构造*

三角网格构造是逆向工程CAD建模技术中的一个重要步骤。提出了一种针对扫描线点云数据的三角网格构造方法,该方法的核心思想是合并同层的扫描线,并且标记合并后扫描线的间断点,以便三角网格构造和拓扑结构建立。该方法解决了对于包含分叉和内孔特征的三角网格的构造问题。     

Spanning Forests of a Digraph and Their Applications

We study spanning diverging forests of a digraph and related matrices. It is shown that the normalized matrix of out forests of a digraph coincides with the transition matrix in a specific observation model for Markov chains related to the digraph. Expression are given for the Moore–Penrose generalized inverse and the group inverse of the Kirchhoff matrix. These expressions involve the matrix of maximum out forest of the digraph. Every matrix of out forests with a fixed number of arcs and the normalized matrix of out forests are represented as polynomials of the Kirchhoff matrix; with the help of these identities new proofs are given for the matrix-forest theorem and some other statements. A connection is specified between the forest dimension of a digraph and the degree of an annihilating polynomial for the Kirchhoff matrix. Some accessibility measures for digraph vertices are considered. These are based on the enumeration of spanning forests.     

The bondage number in complete t-partite digraphs

The bondage number of a digraph D is the minimum number of arcs whose removal results in a new digraph with larger domination number. In this paper, we determine the bondage number in complete t-partite digraph .     

The bondage numbers of extended de Bruijn and Kautz digraphs

In this paper, we consider the bondage number b(G) for a digraph G, which is defined as the minimum number of edges whose removal results in a new digraph with larger domination number. This parameter measures to some extent the robustness of an interconnection network with respect to link failures. By constructing a family of minimum dominating sets, we compute the bondage numbers of the extended de Bruijn digraph and the extended Kautz digraph. As special cases, we obtain for the de Bruijn digraph B(d, n) and the Kautz digraph K(d, n) that b(B(d, n)) = d if n is odd and d ⩽ b(B(d, n)) < 2d if n is even, and b(K(d, n)) = d + 1.     

Möbius–deBruijn: The product of Möbius cube and deBruijn digraph

Möbius cube and deBruijn digraph have been proved to be two of the most popular interconnection architectures, due to their desirable properties. Some of the attractive properties of one, however, are not found in the other. The Möbius–deBruijn architecture, proposed in this paper, is the product of Möbius cube and deBruijn digraph, which is a combination of the two architectures. It employs the Möbius cube as a unit cluster and connects many such clusters by means of given number of parallel deBruijn digraphs. Consequently, the Möbius–deBruijn provides some of the desirable properties of both the architectures, such as the flexibility in terms of embedding of parallel algorithms, the high level of fault-tolerant, and the efficient inter-cluster communication. The proposed architecture also possesses the logarithmic diameter, the optimal connectivity, and the simple routing mechanism amenable to network faults. The methodology to construct the Möbius–deBruijn can apply to the product of deBruijn digraph and other hypercube-like networks, and also applies to the product of Kautz digraph and hypercube-like networks.     

On the k-tuple domination of de Bruijn and Kautz digraphs

In a digraph G, a vertex u is said to dominate itself and vertices v such that (u,v) is an arc of G. For a positive integer k, a k-tuple dominating set D of a digraph is a subset of vertices such that every vertex is dominated by at least k vertices in D. The k-tuple domination number of a given digraph is the minimum cardinality of a k-tuple dominating set of the digraph. In this letter, we give the exact values of the k-tuple domination number of de Bruijn and Kautz digraphs.     

Depth-first K-trees and critical path analysis

Summary K-trees are tree representations of digraphs, and depth-first K-trees are Tarjan's depth-first search trees made explicit. Algorithms are given for the construction of depth-first K-trees. The depth-first K-trees are then used as a basis for nonrecursive algorithms for the following: topological ordering of an acyclic digraph, critical path analysis of a scheduling network, determination of longest paths from a single source in an acyclic digraph, and determination of strong components of a digraph. The time complexity of each algorithm is O(m), where m is the number of arcs in the digraph. Storage structures for K-trees are defined, and the updating of a scheduling network is discussed.The research reported here has been supported in part by the National Science Foundation under grants MCS 73-03371 A01 and MCS 77-01462     

基于PCA学习子空间算法的有限汉字识别

采用PCA学习子空间方法来进行灰度图象上字符的识别,不仅克服了传统的基于二值化字符特征提取和识别所带来的主要困难,还尽量多地保存了字符特征,该算法在PCA子空间的基础上,通过反馈监督学习的方法使子空间作旋转调整,从而获得了更好的分类效果,特别当字符类别数不是很大时,子空间的训练时间也将在可接受的范围之内,应用效果也表明,采用PCAA学习子空间算法对车牌汉字这一有限汉字集进行识别,取得了较好的效果,实用价值较高。