|
|
|||||||||||||||||||||||||||||
论文吧首页
| 写作指南
| 示范论文
| 分类地图
| 参考资料
| 论文吧
| 专题
| 留言
| 搜索
|
(1)孤立域:没有父域和子域;
(2)起始域:没有父域,只有一个子域;
(3)终止域:没有子域,只有一个父域;
(4)单一域:只有一个父域和一个子域;
(5)分叉域:有多个子域,至多一个父域;
(6)汇合域:有多个父域,至多一个子域;
(7)交叉域,有多个父域和多个子域。
在图3.a中,孤立域1表示一个图元,即一条线段。在图3.b中,交叉域7连接起始域2和终止域3,组成一条线段,还表示该线段与另一线段相交于此。可以看出,单义域不仅为单一完整图元的获得提供几何数据,而且也为图元之间的拓扑关系提供线索,单义域邻接图比较完整地了表达图象的几何数据与拓扑关系。
4 圆弧识别
4.1 确定种子圆弧
本文采用假设验证法从单义域中提取圆弧域,主要思路是,先假设候选域为圆弧域,采用最小二乘法来计算其几何定义数据(圆心和半径)[8],并计算平均径向误差和最大径向误差,如果小于阈值,则判定为圆弧域,作为种子圆弧,以指导圆弧和圆的识别。
在确保计算精度的前提下,为提高速度,平均抽取五个游程的中点,如果单义域游程数小于五,则全部选取。基于径向误差最小,对样点采用最小二乘法拟合,下面给出计算公式:
设给定样点为
,所求圆心为
,半径为
,平均径向误差为
,最大径向误差为
,则

,
,
,
,
。
4.2 识别圆弧与圆
在单义域邻接图中,从顶点中选取圆弧域,作为种子圆弧,以此为起始点,遍历得其邻接域(或邻接域的邻接域),如果属于同一圆,则种子圆弧生长,并继续遍历,否则终止,即可得一弧,如果首末闭合,则得一圆。圆弧及圆的识别算法如下所述:
(1)从单义域邻接图的顶点集中提取弧形域,作为种子圆弧,设为当前域;
(2)如果当前域无邻接域,则转到(4);
(3)取当前域的邻接域i,i从1到n,n为邻接域总数
{
如果邻接域i与种子圆弧同圆,则种子圆弧生长,并设邻接域i为当前域,同时返回(2);
取邻接域i的邻接域j,j从1到m,m为邻接域i的邻接域总数
{
如果邻接域j与种子圆弧同圆,则种子圆弧生长,并设邻接域j为当前域,同时返回(2);
}
}
(4)如果该路径为回路,则得一圆,否则得一圆弧,计算其几何数据,获得矢量表达。
本算法不仅能够识别单独的圆弧和圆,而且还兼顾圆弧(圆)与线段、圆弧(圆)的相交和相切等情况。在图4中,原始图象(a)经过图文分离和粗细分离,包含圆和多种圆弧,(b)对图象多义域进行分裂,(c)为图象的单义域表达,(d)为矢量化图形。由结果可以看出,该法识别能力很强,能处理多种圆弧和圆。

5 结束语
本文介绍了一种基于单义域邻接图的圆弧与圆识别算法,与以前方法相比,无需线段逼近和模式匹配,可以直接提取圆弧,对短小圆弧也有较高识别率,对圆弧(圆)与线段、圆弧(圆)的相交和相切等也同样适用,该方法已被应用于我们开发的工程图纸扫描图象识别与理解系统之中,效果较好。但,仍需进一步完善,研究各种复杂情况,以提高识别范围。单义域邻接图的描述模型也可用于线段完整识别及字符识别等方面。
参考文献
[1] Vijay Nagasamy and Noshir A. Langrana. Engineering Drawing Processing and Vectorization System. Computer Vision, Graphics, and Image Processing, 1990,49:379-397
[2] Dov Dori, Member IEEE. Vector-Based Arc Segmentation in the Machine Drawing Understanding System Environment. IEEE Transactions on Pattern and Machine Intelligence, 1995,17(11):1057-1068
[3] 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. 大连理工大学博士论文,1998,3
[4] C.-C. Han and K.-C. Fan. Skeleton Generation of Engineering Drawings via Contour Matching. Pattern Recognition ,1994,27(2): 261-275
[5] 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
[6] 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
[7] 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
[8] 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
An Algorithm for Recognizing Circular Arcs and Circles Using Primitive Region Adjacency Graph
Abstract The scanning input and recognition of engineering drawings is a key step in CAD, and is to reuse lots of engineering drawings. In study on recognition of scanned image of engineering drawings, the recognition for circular arcs is an important and difficult problem. Recent algorithms of recognizing arcs are mainly about approximation with lines. This paper presents an algorithm for recognizing arcs and circles using Primitive Regions Adjacent Graph. The method can directly extract arc. The binary image is encoded with black horizontal runlength. A stripe region consists of correlative runlengths with the same width and topology. The stripe regions then can be segmented as some primitive regions (line and arc). The Graph is used to describe geometrical property and topological constraint. The primitive region supplies shape information (line, arc, arrow etc) improving integrality of recognition. After extraction of the arc region from regions, the seed for an arc is obtained. By traversals for the graph, the seed arc grows by constrains for the same circle. Some applications to recognize arcs and circles are finally provided, which show that the algorithm is effective and robust, can solve well intersection and tangency of between a arc and a line or a arc.
Key words engineering drawings, vectorization, recognition for circular arc, stripe region, Primitive Region Adjacency Graph
转贴于 大学生素质网 Http://Www.Uscc.Com.Cn
本文引用地址:http://www.uscc.com.cn/Discourse/Demonstrate/Science/Engineering/200609/Discourse_5101.htm