摘要
提出了对扫描图像中直线的一种分类识别算法,该算法不必经过细化直接对直线进行识别且对交叉区域处理效果明显。本文首先提取图像的由梯形块表示的几伺信息,然后由分类后的梯形块生成若干链表,最后按类识别。该算法不必单独处理线的交叉问题。实验结果令人满意。
A new approach for the recognition of lines in scanned engineering drawings is proposed in this paper. This method executes without thinning and deals with intersection areas effectively. In this paper, geometric information, which is composed of trapezoid black blocks, is extracted from drawing image and classified and discussed in detail. The classified trapezoid black blocks are organized into several chains later, and then straight lines are generated based on the chains. The algorithm modifies only trapezoid black blocks that placed in intersection areas in recognition process instead of dealing with intersection points after recognition. The results demonstrate the effectiveness of this method.
出处
《模式识别与人工智能》
EI
CSCD
北大核心
1999年第2期234-240,共7页
Pattern Recognition and Artificial Intelligence