图纸矢量化中的线弧分离与识别

摘要

As an important branch in modern image processing, drawing imagevectorization has high research value and broad applications. In this paper, a general method for meta-graph recognition is proposed based on curve fitting and separation of line and smooth curve. Though line and arc separation, sharp angle and arc end judgement, and segmentation of inverted tangency arcs, various basic metas can be well recognized. Both theoretical analysis and experiment results have shown that the proposed method has good and efficient performance for different drawing images.%图纸矢量化是现代图像图形处理技术发展的一个重要分枝，具有很高的研究价值及广泛的应用前景。本文针对图纸矢量化中的图元识别技术，提出了系统的基于线弧分离和拟合识别方法。通过长直线和圆弧分离、尖角与弧结束判定及反向相切弧分离，进而识别多种基本图元。理论分析和实验结果均表明，该方法对于不同图纸中线弧图元可以进行有效的识别。