摘要
初步实现了一个工程图纸的自动输入与识别系统。该系统把工程图纸的扫描数字化图象进行矢量化,转换成CAD系统可以接受的文件格式。研究了扫描图象的预处理和矢量化方法:预处理包括图纸扫描输入、图象二值化、图象平滑与去噪、线条细化和曲线跟踪;矢量化采用直线拟合方法。提出了一种改进Hilditch细化算法并给出了程序实现方法,有效地减小了图像骨架偏移和毛刺对细化结果的影响。该文开发的系统可以识别含有直线、圆和圆弧等简单图形的工程图纸。
This paper implements an experimental system for automatically inputt in g and recognizing engineering drawings.The system transfers the scan digitized image of engineering drawings to the DXF graphics file format in CAD system.The methods for preprocessing and vectorizing the scan image are described in the p aper.The preprocessing steps include drawing scanning,binary image transformin g,line thinning and cure tracing.The vectorizing uses straight-line fitting m ethod.An improved Hilditch thinning algorithm is presented in this paper.The n ew method makes use of eight templates to traverse the whole image before every iteration in the iterative procedure of the Hilditch algorithm.This method is more efficient than Hilditch algorithm because it is better to generate a new s keleton than to remove the current one and to trace a new one in the next itera tion.The performance of the system is evaluated with several test drawings and the experimental results are acceptable.
出处
《计算机工程与应用》
CSCD
北大核心
2003年第33期138-140,共3页
Computer Engineering and Applications
关键词
图象识别
图象二值化
细化
直线拟合
矢量化
image recognition,binary im age,thinning,straight-line fitting,vectorization