摘要
目的提高直线识别的准确度和精度,更好地解决线图形的直线矢量化中存在的不足,如细、斜线矢量化效果差,直线易断裂,端点定位不准等.进一步推进工程图纸理解系统的发展.方法依据工程图纸中直线像素群体出现的特征,搜索符合某群体特征的像素群,通过采集试探点组信息来统计判定直线是否可以定向生长,并采用节点窗口定位法提高端点定位准确度.结果实验对比表明:笔者方法受线宽的影响小,增强了抵抗直线断裂的能力,对提高直线识别的质量是有效的.结论基于像素群体特征,结合统计判定直线生长的方法,并在节点定位时适时引入节点窗口定位,使得此方法在直线识别上具有一定的速度与精度优势.
In the field of line drawing vectorization, searches about line vectorization are more important than other factors, such as vectorization of arc and ellipse. However, it is still difficult to avoid or reduce certain disdvantages, for example, the unfavorable vectorization of thin and slope lines, the easily-broken lines, and questions about end-points bad location. This paper proposes a new algorithm of line recognition based on character of pixels. It searches a group of pixels having some special characters at first, then judges if the line can grow by a statistical way called one group of three scanning lines, and finally uses a window to locate its end-points to improve its precision. Experiments show that this method enhances the anti-noise performance and decreases the quantity of fragmentation. Thus, this algorithm is valid.
出处
《沈阳建筑大学学报（自然科学版）》
EI
CAS
2007年第1期147-150,共4页
Journal of Shenyang Jianzhu University：Natural Science
基金
国家科技成果重点推广计划项目(2004EC000096)
关键词
工程图纸
直线识别
矢量化
稀疏像素跟踪
engineering drawing
line recognition
vectorization
sparse pixel tracking
