【作者】 姬忠校；

【导师】 董惠；

【作者基本信息】 西安建筑科技大学 ， 控制理论与控制工程， 2008， 硕士

【摘要】 机械图纸矢量化是将机械图纸经扫描设备输入计算机中的位图图像进行处理、分析和识别,并最终将其转化为计算机辅助设计软件可以识别的矢量图形文件格式的过程。它涉及图像处理与分析、人工智能和模式识别等领域的技术,是一种具有重要理论意义和巨大商业价值的技术;图纸拼接技术可以将分批扫描输入的小图纸拼接成一幅大图纸,是一种非常实用的技术。本文主要对图纸矢量化过程和图纸拼接技术展开讨论和研究。图纸矢量化过程主要包括图像二值化、图像去噪和平滑、图像细化、图像识别与矢量化以及矢量化输出。本文在分析数学形态学运算特点的基础上,对基于数学形态学的混合去噪方法进行了改进和扩展,这种改进后的方法可以保障图像的完整性和准确性,且具有较好的去噪效果;直线、圆和圆弧是图纸中的基本图像元素,是图像识别与矢量化的主要处理对象,本文采用Hough变换的方法检测直线,并提出通过截取直线上一段无交点的线段并对这条线段进行Radon变换来确定直线线宽的方法,这种方法确定的线宽信息较为准确;通过分析圆和圆弧的特点,文中得出了一种满足随机Hough变换边界条件的网格图像搜索算法,通过对搜索出的网格图像进行随机Hough变换确定圆和圆弧的参数,使得Hough变换的计算量大大减少;文中通过将矢量化结果按照AutoCAD系统的DXF文件格式进行记录,实现了矢量化输出的过程。根据图纸图像的特点,本文采用基于图像模板的局部匹配方法。它通过截取重叠区域的图像作为模板在另一幅图像中寻找最佳匹配位置,从而确定出图像的拼接参数,实现图像的拼接过程。文中采用Matlab软件对改进的基于数学形态学的混合去噪方法、Radon变换确定线宽的方法和图像拼接过程进行仿真和验证,效果良好。更多还原

【Abstract】 Machine Drawing Vectorization is the process of processing, analyzing and recognizing the scanned bitmap image of a machine drawing, which is input to computer through scanner. Finally, the result of vectorization is converted into vector format so as to be recognized by Computer Aided Design software. The technique the paper has applied is referred to image processing and analyzing, artificial intelligence, pattern recognition etc, and has theory significance as well as large commercial value. The technique of Image Mosaic may combine these pieces of the scanned drawing into a whole drawing image. So it is useful in the real life. The paper mainly discusses and researches on the technique of Machine Drawing Vectorization and Image Mosaic.The process of Machine Drawing Vectorization mainly includes image binarizing, noise-reducing and smoothing, image thinning, image recognizing, vectorizing and its exporting. Through analyzing the operation of mathematical morphology, the method of combination noise-reducing based on mathematical morphology is improved and expanded in the paper. This improved method can ensure the integrity and accuracy of image and has the better effect of noise-reducing. The basic image elements which are line, circle and arc are mainly objects of image recognizing and vectorizing. In the paper, Hough transformation is used for detecting line. The method of exactly acquiring the width of line is proposed. It applies Radon transformation to the line segment which is intercepted from the whole line and hasn’t the intersection in. According to the characters of circle and arc, it is introduced in the paper that the gridding image seeking method which meets the conditions of random Hough transformation. Further the parameters of circle and arc result from the gridding image’s random Hough transformation. This method reduces the complexity of Hough transformation. The process of vectorizing output is realized through the result of vectorizing image being recorded in the format of DXF file in AutoCAD software.According to the characters of the machine drawing image, the part matching method is used, based on the template of image. The template the method used is obtained from the overlap of images, and then the optimal mosaic parameters is worked out through the template seeking for an optimal matching position in the other image.Finally, the simulation and tests have verified the improved method of combination noise-reducing based on mathematical morphology, the line width with Radon transformation, as well as the method of Image Mosaic with MatLab Tools.更多还原