基于Freeman链码的直线识别方法  

Freeman Encoding-based Line-segments Recognition

作　　者：王平[1] 董玉德[1] 罗喆帅[1] 

机构地区：[1]中国科学技术大学计算机科学与技术系,合肥230027

出　　处：《计算机工程》2005年第10期171-173,199,共4页Computer Engineering

基　　金：安徽省自然科学基金资助项目(010042209);中国科学技术大学青年科学基金资助项目(99200)

摘　　要：直线是工程图纸中出现频率最多的图元,许多更高层次的图元都是以直线段为基础而得到的,因此直线段的识别在图纸矢量化研究中占有重要地位。针对细化后的扫描图,该文将Freeman针对直线编码提出的3个约束条件应用到直线的识别之中,取得了较好的效果。通过两次直线段的合并操作较好地解决了短线段过多的问题,并且恢复了在图纸中占有重要地位的线宽信息。Straight-line segments are the most prominent constituent in most engineering drawings. And many high level constituent parts can be generated through line-segments approximation, so recognition for line-segments is the most important stage in vectorization. Three restrictions about Freeman encoding for line are applied to the recognition for straight-line segments in the thinned scanning-drawings. The overabundance of short-line segments is solved through two line-merging procedures and the line-width is also extracted.

关 键 词：工程图纸 直线识别 线宽信息恢复 直线段合并 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]