基于单义域邻接图的扫描工程图样自组织智能识别理论与技术研究
详细信息    本馆镜像全文|  推荐本文 |  |   获取CNKI官网全文
摘要
扫描工程图样识别是图象处理、模式识别和人工智能等多种学科的综
    合应用,直接面向企业需求,具有很高的理论意义和应用价值,是CAD
    领域的重要课题。经过多年研究,扫描工程图样识别已经取得较大进展,
    部分实现象素到矢量的转换。但是,现有识别方法多拘泥局部,串行处理,
    已实现的识别能力与质量离实际需求还有很大差距,识别理论和方法有待
    突破。
     工程图样是工程图元的有机集合,可看作线条图形,而扫描工程图样
    是象素的自然集合。为将象素聚合为工程图元,本文力求加强表达单元的
    整体性,提高表达层次,重视各种关联,根据启发信息选择和组织识别数
    据和知识,进行分层次处理和自组织推理。本文提出一种称为单义域的新
    的图象表达单元,先将象素矩阵转化为单义域及其拓扑关系的集合。同一
    层次数据之间相互关联。不同层次数据之间也相互关联。识别是一个反复
    自组织的推理过程。本文提出基于单义域邻接图的扫描工程图样自组织智
    能识别。在算法实现上,采用模糊分类、遗传算法和面向对象知识表示等。
    处理过程分为三个阶段:
     (1)提取扫描工程图样的线条特征,构建单义域邻接图来表达形状
     与拓扑信息,统一表达字符和图形的结构特征,使后续的识别
     处理能在较高层次的基本单元上进行。
     (2)遍历单义域邻接图,提取字符及其笔划特征,提取完整的几何
     图元,采用矢量邻接图来组织获取的信息。
     (3)基于矢量邻接图,采用面向对象知识表示来组织矢量之间的约
     束知识,组合工程图元,同时提取关联信息,从而构建工程图
     元邻接图。
     在识别单义域、矢量和工程图元时,既注重横向的关联又重视纵向的
    关联。横向关联指图面不同部分之间的直接关联,纵向关联指单义域和它
    所构成的高层工程图元之间的关联。在识别中,先获得部分特征再生成整
    体结构,然后又用整体结构去指导部分特征的进一步把握;高层信息从低
    层数据获得,反过来又去指导低层数据,不同层次之间相互作用。在智能
    推理中,根据启发信息自动选择相应的识别知识,同时不断调整识别参数,
    以适应不断的数据变化。自组织识别在关联数据相互依赖、相互影响和相
    互作用的协作中进行,错综复杂的关系相互协调完成整个识别。
     工程图样包含的图形和字符均可看作线条,线条之间存在多种连接关
    系。扫描工程图样识别先要获取图象的线条及其关系表达。本文所提出的
    单义域表达单元扩大了连通域表达范围,包括线段、圆弧、箭头和交点。
    
    
    采用游程邻接图表达二值图象,然后作深度优先遍历,基于游程宽度和拓
    扑一致形成条形域。引人模糊逻辑对条形域进行分类,获取初步矢量信息,
    对其中多义域做单义分裂,以线段和圆弧为基元,采用遗传算法来实现。
    单义域是具有矢量特征的局部象素合理聚集,反映关联象素的整体特性。
    继承游程的拓扑关系,构建单义域邻接图。基于单义域邻接图自组织识别
    扫描工程图样,在处理效率和抑制噪音误差影响等方面更为优越。
     在单义域邻接图基础上,可对字符和图形进行自组织识别,在字符笔
    划域基础上进行字符提取,同时提取其笔划特征,为将来字符识别提供结
    构信息。根据字符域大小对字符域外接矩形进行自适应膨胀。根据字符域
    膨胀矩形相交来判定字符邻近程度,再加上字符共线为判据来生成字符串
    域。利用同串字符的外接矩形中心和所附图形对字符进行定向。在提取线“
    段、圆弧和圆时,先从种子域线段或圆弧出发,按照同线或同圆的要求识
    别处理,进行邻接图深度捏索,种子矢量不断生长,几何参数不断调整,
    从而获得完整信息。提取的信息采用矢量邻接图来组织。
     上述工作可以实现多种扫描工程图样中的字符、线段、圆弧、圆和箭
    头的提取。但是,工程图样是工程图元的集合。本文采用面向对象方法,
    对工程图元进行对象设计。给出基于矢量邻接图的工程图元识别方法,根
    据已经提取的几何图元和字符信息,从某一特征图元出发,选择相应的识
    别知识(工程图元组成语法),搜索所有其它组元,提取完整信息。文中
    分别提取点划线和虚线的线段、圆弧和圆,还提取剖面线和尺寸,同时与
    约束图元关联。
     上述识别方法已在开发的扫描工程图样识别原型系统中实现,软件采 一
    用面向对象和过程技术分析和设计。基于单义域邻接图的自组织识别方法
    丰富了扫描工程图样识别方法,加强宏观和整体处理能力,利用了更多的
    关联,力求对识别数据和知识进行自组织。对多种扫描工程图样进行识别,
    效果较好。
Recognition of scanned engineering drawings is a comprehensive application
     that relates to multi-discipline, such as image processing, pattern recognition, and
     artificial intelligence, etc. The research on recognition is important in theory and
     practical application. The recognition is one of key issues in the field of CAD.
     Many progress in the area have been made, raster images can be partly transformed
     into vector data that can be used in a CAD system. However, many current
     approaches are limited on capturing local features and process sequentially, and the
     results recognized are not satisfactory, new approaches of recognition still have to
     be researched.
     An engineering drawing is a set of associated entities, which are line-like, and
     a scanned engineering drawing is composed of pixels. To extract entities from
     pixels, recognition process should be self-organizationally implemented by levels.
     The approach developed by author aims to capture more global features and
     intelligently infer with combining lower and higher local infonnation. The
     recognition approach places more emphasis on association relations among features.
     The data and knowledge are chosen and organized according to contexts of data.
     The recognition is to transform pixels into primitive regions. The features in the
     same level are associated each other. The features in different levels are also
     hierarchically associated each other. The recognition is to?infer self-
     organizationally using those associations. Scanned engineering drawings are self-
     organizationally recognized based on Primitive Region Adjacency Graph (PRAG).
     Author uses fuzzy classification, genetic algorithm and object-oriented knowledge
     representing in developing algorithms. The steps of processing is as follows:
     (1) To extract stripe features from an image, a PRAG is used to represent
     geometrical and topological data. It represents structure features of
     characters and graphics, and provides primary data for later processing.
     (2) Characters and their strokes, and integrated graphic primitives are
     extracted form the PRAG. Then, vectors recognized are stored in a vector
     adjacency graph (VAG).
     (3) Based on the VAG, constraint knowledge among vectors is organized
     using object-oriented knowledge representing. Entities are extracted, and
     associations among them are also extracted. The drawings are represented
     using an entity adjacency graph (EAG).
     Primitive regions, vectors and entities are extracted using association of
     features. In the same level, the global structure is constructed with local features
     Ill
    
    
    
    
    
    
    
    
    
     extracted. Then, it is to guild to capture local features. The recognition uses
     association of different levels. Knowlcdge of recognition is chosen automatically
     according to the start data, and some thresholds are adjusted during inferring to
     adapt to context changing. The associations depend on and affect each other. The
     association relations should be self-organized during the recognition
     implementation.
     Components of graphics and texts in drawings can be viewed as stripe regions.
     Recognition of images is to get stripes and their relationships. A new primitive
     region structure and a adjacency graph are developed for representing a scanned
     drawing. A primitive region can represent a line, an arc, an arrow or an intersection
     block, which enlarge the scope of quadrangle-like regions. A binary image is
     represented using a mn-length adjacency g
引文
[1] 大连理工大学工程画教研室编.机械制图.北京:高等教育出版社,1995,4:1—2
    [2] 张李荪.工程图纸扫描输入及矢量化在水利工程设计中的应用.江西水利科技,1997,23(8):155—157
    [3] 孙家广,杨长贵编著.计算机图形学.北京:清华大学出版社,1994,5
    [4] 蔡青,高光焘主编.CAD/CAM 技术的可视化网络化集成化智能化,西北工业大学出版社,1996
    [5] Rik D. Janssen and Albert M. Vossepoel. Adaptive Vectorization of Line Drawing Images. Computer Vision and Image Understanding, 1997, 1,65(1): 38—56
    [6] 周桂英.扫描仪的技术现状及市场趋势.电子计算机与外部设备,1999,9,23(5):62—64
    [7] Heinrich bley. Segmentation and Preprocessing of Electrical Schematics Using Picture Graphs. Computer Vision, Graphics, and Image Processing,1984, 28:271—288
    [8] J.E. Den Hartog, T. K. Ten Kate and J. J. Gerbrands. Knownledge-Based Interpretation of Utility Maps, Computer Vision and Image Understanding,1996, 63(1): 105—117
    [9] Jodeph SH, Pridmore TP. Knowledge Directed Interpretation of Mechanical Engineering Drawings. IEEE Trans on PAMI, 1992, 14 (9): 928—940
    [10] 闵卫东,唐泽圣,唐龙.工程图尺寸标注识别的研究与识别.计算机学报,1994,17(1)
    [11] 李宾,谭建荣,彭群生.工程图扫描图象中剖面区域的整体识别.工程图学学报,1996,2
    [12] 高玮,谭建荣,彭群生.基于图形约束的工程图纸扫描图象表面粗糙度的自动识别.计算机学报,1996,19(6)
    [13] Atish K. Das and Noshir A. Langrana. Recognition and Integration of Dimension Sets in Vectorized Engineering Drawings. Computer Vision and Understanding, 1997, 68(1): 90-108
    [14] 吴成东,宣国荣.工程三视图的三维自动恢复,中国智能CAD'94,清华大学出版社,1994:40—46
    [15] 高玮,彭群生.基于二维视图特征的三维重建.计算机学报,1999,5,22(5):481—485
    [16] 陆国栋.基于工程语义的三维重建研究.(博士学位论文).浙江,浙江大学,2000,1
    
    
    [17] Dov Dori and Karl Tombre. From Engineering Drawings to 3D CAD Models: are We Ready now?. Computer-Aided Design, 1995, 4, 27 (4): 243—254
    [18] 王金鹤.基于条形域结构的扫描图样识别理论与方法研究(博士学位论文).大连,大连理工大学,1999,6
    [19] 谭建荣,彭群生.工程图扫描图象的整体识别方法.谭建荣主编.计算机工程图学的操作与实践,北京:电子工业出版社,1994
    [20] 李伟青,谭建荣,彭群生.基于图段结构的整体识别方法的研究.计算机学报,1998,21(8):753—758
    [21] 李宾.基于认知模型的工程图层次矢量化算法研究(博士学位论文).浙江,浙江大学,1998,6
    [22] Lousia Lam, Seong-Whan Lee, and Ching Y. Suen. Thinning Methodologies —A Comprehensive Survey. IEEE Transactions on Pattern Analysis and Machine Intelligence, 1992, 14(9): 869—885
    [23] Vijay Nagasarny and Noshir A. Langrana. Engineering Drawing Processing and Vectorization System. Computer Vision, Graphics, and Image Processing,1990, 49: 379—397
    [24] 霍宇翔,丁宇,陈耘等.细化畸变节点形态分析及修正策略研究.计算机辅助设计与图形学学报,1997年11月,9(6):500—504
    [25] Dov Dori, Yubin Liang, Joseph Dowelll,and Lan Chal. Sparse-pixel Recognition of Primitives in Engineering Drawings. Machine Vision and Applications, 1993, 6:69—82
    [26] Dov Dori. Vector-Based Arc Segmentation in the Machine Drawing Understanding System Environment. IEEE Transactions on Pattern and Machine Intelligence, 1995, 17(11): 1057—1068
    [27] Dov Dori and Yelena Velkovitch. Segmentation and Recognition of Dimensioning Text from Engineering Drawings. Computer Vision and Image Understanding, 1998, 2, 69(2): 196—201
    [28] 王健,张树生,王广.基于节点的线状图骨架提取算法研究.计算机研究与发展,1999,6,36(6):725—731
    [29] C.-C. Han and K.-C. Fan. Skeleton Generation of Engineering Drawings via Contour Matching. Pattern Recognition ,1994, 27(2): 261-275
    [30] 颜学阶,常明,朱林.基于线性轮廓匹配的工程图识别算法研究.华中理工大学学报,1997年3月,25(3):34—37
    [31] Theo Pavlidis. A Vectorizer and Feature Extractor for Document Recognition.Computer Vision, Graphics, and Image Processing, 1986, 35:111—127
    [32] 李宾,谭建荣,彭群生.一个基于扫描串的统一整体矢量化算法.软件学报,1998,6,9(6):426—431
    [33] 周辉.扫描工程图纸识别输入处理与联机手绘图形输入技术的研究(博士学位论文).大连理工大学,1998,3
    
    
    [34]徐建铿,高国安.基于TAG的工程图纸矢量化算法,中国图象图形学报,1996,1(5,6):457—460
    [35]邹荣金,陈冀兵,苏峰等.基于特征段的正交矢量化及其符号识别方法.计算机研究与发展,1999,1,36(1):85—90
    [36]邹荣金,蔡士杰,张福炎等.字符粘连及字线相交的分割与识别方法.软件学报,1999,10(3):241—247
    [37]胡友兰,黄树槐,常明.工程图中字符分离和标注字符串生成技术.华中理工大学学报,1997,25(3):30—33
    [38]陈勇,朱林,常明.工程图中粘连字符的提取与分割.华中理工大学学报,1996,24(4):23—26
    [39]张德喜.手写体汉字机器识别技术的现状分析.许昌师专学报,1999,3,18(3):91—95
    [40]刘文印,唐龙,唐泽圣.Dov Dori,一种在矢量基础上进行图形识别的通用方法,软件学报,1997,8(5)
    [41]Noshir A. Langrana, Yuan Chen, and Atish K. Das. Feature Identification from Vectorized Mechanical Drawings, Computer Vision and Image Understanding, 1997, 68(2)
    [42]Chan Pyng Lai and Rangachar Kastutri. Detection of Dimension Sets in Engineering Drawings. IEEE Transactions on Pattern Analysis and Machine Intelligence, 1994, 16(8): 848—855
    [43]刘世霞,胡事民,汪国平等.基于三视图的三维形体重建技术.计算机学报,2000,2,Vol.23 No.2,141—146
    [44]陆国栋,阮建中,江涛等.基于工程图样的三维重建技术研究与实现.计算机辅助设计与图形学学报,1999,11,Vol,11 No.6,516—520
    [45]刘薇.光栅矢量化软件评测.计算机世界,1998年8月17日,E11,E13
    [46]李新友,唐泽圣,张凤昌等.TH-DAIMS3.0清华图纸自动输入与管理系统.计算机辅助设计与图形学学报,1997,9,9(5):450—454
    [47]王德英,李新友,唐泽圣等.图纸图象上线条的交互拾取.计算机学报,1997,10,20(10):925—932
    [48]尧晓华,朱建新,明正勤.工程图的快速半自动矢量化和点阵拾取算法.华中理工大学学报,1997,8,25(8):56—58
    [49]李欣.泵水利模型数字化CAD系统—工程图中的曲线识别与数字化的研究(硕士学位论文).1998,6,大连理工大学
    [50]徐尧东,张忠苗,岑岗.工程图扫描输入与识别技术.计算机辅助工程,1997年4月:44—51
    [51]胡友兰,常明,田亚梅等.工程图纸智能识别的分层理解模型.计算机工程,1998,12,24(12):50—55
    [52]周昌乐著.手写汉字的机器识别.北京:科学出版社,1997,8
    [53]王伟编著.人工神经网络原理—入门与应用.北京:北京航空航天大学出版社,1995,4
    
    
    [54]陈振宇.基于图形整体拓扑特征的工程图纸扫描图象自动识别输入方法的研究(博士学位论文).大连,大连理工大学,1998,6
    [55]郑南宁著.计算机视觉与模式识别.北京:国防工业出版社;1998,3:160—168
    [56]高文,陈熙霖.计算机视觉—算法与系统原理.北京:清华大学出版社,广西科学技术出版社,1999,2
    [57]王润生.图象理解.国防工业出版社,1994
    [58]刘敬军,张申生.综合面向对象和面向过程的对象过程方法.计算机工程,1998,4,24(4):13—15
    [59]张习文,欧宗瑛.工程图纸扫描图象的单义域邻接图描述及其在矢量化中的应用,中国图象图形学会东北分会第四届会议论文集,1999.8,大连理工大学,130—133
    [60]张习文,欧宗瑛.基于单义域邻接图的圆弧与圆识别,中国图象图形学报,2000.1:70—74
    [61]金连文,徐秉铮.手写体汉字识别中的一种新的特征提取方法——弹性网格方向分解特征.电路与系统学报,1997,8,2(3):7—12
    [62]L. Y. Tseng and C. T. Chuang. An efficient Knowledge-Based Stroke Extraction Method for Multi-Font Chinese Characters. Pattern Recognition,1992, 25(12): 1445-1458
    [63]S. Di Zenzo, L. Cinque, and S. Levialdi. Run-Based Algorithms for Binary Image Analysis and Processing. IEEE Transactions on Pattern Analysis and Machine Intelligence, 1996, 18(1): 83—89
    [64]姚敏著.计算机模糊信息处理技术.上海,上海科学技术文献出版社,1999,4
    [65]张文景,许晓鸣,丁国骏等.一种基于曲率提取轮廓特征点的方法.上海交通大学学报,1999,33(5):592—595
    [66]王晨,皇甫堪,万建伟等.基于多尺度小波变换的二维图象角点检测技术.国防科技大学学报,1999,21(2):46—49
    [67]周辉,李涛,邢启江等.数字曲线的线性逼近和分段识别.大连理工大学学报,1997,37(5):576—580
    [68]王明江,唐璞山.基于矢量斜率的分段线性拟合.软件学报,1999,10(2):165—169
    [69]Robert Bergevin and Marielle Mokhtari. Multiscale Contour Segmentation and Approximation: An Algorithm Based on the Geometry of Regular Inscribed Polygons. Computer Vision and Image Understanding, 1998, 71(1) :55—73
    [70]Wenhua Wan and Jose A. Ventura. Segmentation of Planar, Curves into Straight Line Segments and Elliptical Arcs. Graphical Models and Image Processing,1997,59(6)484—494
    
    
    [71]艾丽蓉,何华灿.遗传算法综述.计算机应用研究,1997年4月,3—6
    [72]郭子华,庄镇泉.基于遗传算法的地图矢量化算法.中国科学技术大学学报,1998,8,28(4):476—481
    [73]Lakhmi C.Jain Editor. Soft Computing Techniques in Knowledge-based Intelligent Engineering Systems Approaches and Applications. A SpringerVerlag Company, 1997
    [74]王金鹤,欧宗瑛,夏晓东.工程图扫描图象中交叉区域识别处理方法的研究.中国图象图形学报,1999年9月,Vol.4(A),No.9,751—755
    [75]何青,杜永祚,宋之平.一种实用的不变矩计算方法.华北电力大学学报,1998,10,25(4):80—83
    [76]江早,刘积仁,刘晋军.工程图纸图象图文自动分割工具SegChar.软件学报,1999,10(6):589—594
    [77]李伟青,彭群尘.一种新的字符提取和组合算法.工程图学学报,1997,No.2-3:38—45
    [78]冯笑,罗长印,马义德等.从静态字符骨架中提取完整笔划.兰州大学学报,1998,34(1):41—4
    [79]宋晓宇,王永会.直线矢量化算法的研究.沈阳建筑工程学院学报,1997年4月,13(2),105—110
    [80]王金鹤,欧宗瑛,夏晓东.工程扫描图象的直线整体识别算法.中国图象图形学报,1998,11,3(11):912—917
    [81]谭建荣,彭群生.基于图形约束的工程图扫描图象直线整体识别方法.计算机学报,1994,17(8):561—569
    [82]张志鸿,童秉枢.工程图识别中实现图素完整拼接的索引表方法.工程图学学报,1996,1:55—59
    [83]卢朝阳,应道宁,王尔健等.直线、圆弧的矢量化及识别—自适应多重决策算法.潘云鹤主编.中国智能CAD'94,清华大学出版社,广西科学技术出版社,1994,10:58—61
    [84]吴仲科,焦海星等.一种线段和圆弧的逼近方法及其在工程图纸矢量化中的应用.计算机辅助设计与图形学学报,1998,10(4):328—332
    [85]宋晓宇,王永会.工程图自动矢量化算法的设计与实现.中国图象图形学报,2000,1,5(1):66—69
    [86]李伟青,彭群生.一种基于模式的圆的识别算法.软件学报,1999,10(2):129—132
    [87]章毓晋编著.图象工程上册图象处理和分析.北京:清华大学出版社,1999,3
    [88]Liu Wenyin and Dov Dori. A Generic Integrated Line Detection Algorithm and Its Object-Process Specification. Computer Vision and Image Understanding, 1998, 6, Vol. 70 No. 3, 420—437
    
    
    [89]郁晓红,李伟青.工程图纸中虚线的提取算法.计算机应用与软件,1999,5:16—19
    [90]唐荣锡编著.CAD/CAM技术.北京:北京航天航空大学出版社,1994,9
    [91]张海军,牛宝峰.基于标注字符的尺寸标注自动识别.河南师范大学学报(自然科学版),1998.8,26(3):102——104
    [92]李伟青,彭群生,谭建荣.工程图纸中尺寸约束信息的识别与提取.谭建荣主编.计算机工程图学的操作与实践,北京:电子工业出版社,1994:407—410
    [93]谭建荣,董玉德.基于图形理解的尺寸环提取算法及其实现.计算机研究与发展,1999,36(2):192—196

© 2004-2018 中国地质图书馆版权所有 京ICP备05064691号 京公网安备11010802017129号

地址:北京市海淀区学院路29号 邮编:100083

电话:办公室:(+86 10)66554848;文献借阅、咨询服务、科技查新:66554700