摘要
针对常规矢量化几何失真较明显,抗噪声干扰能力弱等不足,本文提出了基于骨架点对划进行跟踪,通过设定规则库进行噪声滤除和形状校正的智能矢量化方法。本方法在矢量化各环节分别采用了动态模板定位骨架点;根据典型噪声和失真的知识,设计准则自动进行噪声滤除和形状校正;利用最大距离法进行矢量化等方法,以克服典型的噪声和干扰。由于基于规则的智能化处理方法的采用,可根据需要灵活加入特定形状校正功能。与常规矢量化方法相比,具有拓扑信息完好,几何保真度高,数据简洁,符合人的直观视觉等优点,可广泛用于文字及其它二维图形识别的预处理。
To improve the normal vectorilization algorithm,a new intelligence algorithm of vector acquisition from analog map that uses predicate rules to automatically reduce noises and refine the outline of objects on paper has been proposed in the article.The method is designed to simulate human's action on how to get vector from an image,mixed with the following technology:dynamic template,statistic weighting,Maximum-error-method,a sets of rule based the noise and distortion knowledge extracted from map and CAD-drawing,is applied cooperatively in these main steps of converting raster data into vector data: skeleton point extraction,lines procession,vertex selection,noise reduction and shape correction.Because of the intelligence method especially the set of rules,a new suitable rule to do shape adjustment for a particular map and CAD-drawing can be added in the set of rule easily, and this makes it convenient to automatic editing.Compared with the normal algorithm performance,vector obtained by this way have radical topology information,lower disfigure,and the vector data has low redundancy and little visual difference to original image.The method can be applied to the preprocessing of OCR and other two-dimensional shape recognition.
出处
《小型微型计算机系统》
CSCD
北大核心
1996年第5期6-11,共6页
Journal of Chinese Computer Systems
关键词
地图
工程图纸
智能化处理
计算机
识别
Vector acqusition
Skeleton points extractionAI technology