【作者】 冯奂；

【导师】 张树生；

【作者基本信息】 西北工业大学 ， 航空宇航制造工程， 2006， 硕士

【摘要】 工程图纸识别是图像处理、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,扫描工程图纸识别已经取得较大进展,部分实现像素到矢量的转换。但是,现有的工程图纸识别系统的矢量化结果是以孤立存在的图形基元(直线、圆弧、圆和字符等)为主,这种孤立图元结构的工程图识别结果缺乏矢量之间的逻辑关系、位置关系、拓扑结构关系和工程语义信息等,与高层次的工程图识别理解与智能重用相差较远。要全面实现工程图纸的识别理解与智能重用,完成二维模型到三维模型的转变,就必须在矢量化基础上进一步识别出工程图中的各种图形符号。本文对点划线、粗糙度符号和形位公差符号的识别技术进行了深入的研究。 本文分析了现有的点划线识别算法,指出了其局限性,为了克服现有方法的缺陷,本文根据直线式点划线和圆弧式点划线的构造特点和语法规则,在矢量基础上,提出了一种基于关键图形特征的识别算法。该算法首先提取一条直线或一段圆弧,将它们作为点划线的关键元素:然后根据点划线的语法规则,由关键元素引导,逐步搜索其它构成元素。该算法摆脱了以往仅识别直线式点划线的模式,对圆弧组成的点划线也有了很好的识别,提高了该算法的通用性。针对半径较小的圆或圆弧,本文提出了一种圆的中心线识别算法,以适应点划线为一条直线的特殊情况,并提高了圆的中心线的识别率。 本文还分析了现有的粗糙度符号识别算法,指出了其局限性,为了克服现有方法的缺陷,本文根据机械工程图中粗糙度符号的特点,在矢量基础上,提出并实现了一种基于关键图形特征和标注字符相结合的识别算法。该方法包括3个步骤:首先寻找该类图形的关键元素:然后根据该类图形的具体构造语法规则,由关键元素引导、逐步搜索该类图形的其它结构元素;最后用字符信息来对该类图形的有效性做出判定。该算法除了可以识别直线式粗糙度符号外,还可以对基本粗糙度符号和圆型粗糙度符号进行识别,提高了粗糙疫符号的整体识别率,解决了粗糙度符号另外两种形状的识别问题,由于用字符信息加以辅助判断,该算法的适应性和容错性得到进一步提高,此外该算法还具有识别过程简单、快速等优点。 本文讨论了在工程图矢量化后对形位公差符号的识别工作。针对形位公差自身的特点和语法、语义规则,在矢量基础上,提出并实现了一种基于关键图形特更多还原

【Abstract】 Recognition of engineering drawings concerns the application of multi-disciplines such as image processing, pattern recognition, and artificial intelligence, etc. The research on recognition of engineering drawings is important both on theory and practical application and it is one of key issues in the field of CAD. Many progresses in the area have been made, raster images can be partly transformed into vector data that can be reused in a CAD system. However, the vectorization results made by the existed recognition systems are mainly some isolated graphic units, such as lines, arcs, circles and characters. Such recognition results lack of high-layer logic relations as position, topology and structure relations and engineering semantic information and are far away from the intelligence comprehending and reuse of engineering drawings. Based on the vectorization results, further work should be done to realize better comprehending and reuse for engineering drawings and transform it from 2D to 3D. In this paper, the recognition technology of dash-dotted line, roughness degree symbols and tolerance symbols in shape and relative position are studied.Based on the detailed analysis of the existed dash-dotted line recognition methods, this paper used structure characteristics and syntax rule of line or arc type of dash-dotted line, based on the vector, and presented a recognition algorithm based on key graphics characteristics. This algorithm consists of two main steps: firstly finding a line or an arc, and make them as the key components of dash-dotted line;secondly according to the specific graphic syntax of dash-dotted line, leading by key element, step by step to search other composing elements. This algorithm gets rid of anciently mode which only recognized the line type of dash-dotted line, and has a good result of arc type of dash-dotted line, improved the universal capability. Aim at the arcs or circles which have smaller radius, a recognition algorithm for circle’s centre line is proposed. This algorithm can better adapt to the condition that dash-dotted line was only a straight line, and improved the recognition rate for circle’s centre line.For the roughness degree symbols, this paper studied a new vector based recognition algorithm based on key graphics characteristics and label character according to the structure characteristics and syntax rule of roughness degree symbols, The algorithm consists of three main steps: firstly finding the key component, secondly extending it according to the specific graphic syntax and finally judging the validity by characters. This algorithm can recognize basic roughness and circle type roughness symbol besides line type roughness symbol, and improved the entire recognition rate. With the aide of the assistant judging by characters, powerful adaptability and fault-tolerance ability have been gotten. Experimental results indicate that this algorithm has such advantages as a higher recognition rate, a simpler processing process and a higher speed.Analyzing the characteristics of tolerance symbols in shape and relative position in engineering drawings, a vector-based recognition algorithm based on key graphics characteristics and label character is proposed and implemented. This algorithm consists of three main steps: firstly finding the key component, secondly extending it according to the specific graphic syntax, thirdly judging the validity by characters. If this symbol is very complicated, it can also get the good result by using this algorithm. Experimental results indicate that this algorithm can get a higher recognition rate.All above algorithms mentioned have been programmed by using C++ language in the VC++6.0 system by author. In the practice testing the recognition results are satisfactory.更多还原