基于距离变换和变形校正的点状符号模式识别研究

Research of point symbols recognition based on distance transformation and distortion

针对地图上符号的提取与识别,提出了一种基于距离变换和变形校正的点状符号的模式识别方法。该方法主要针对数值字符小数位和字符变形提出的特殊处理方法,利用了线划弯曲度的概念,根据弯曲密度和线划密度过滤掉非点状符号的线划,然后对字符重心距离的度量达到数值字符小数位的提取,再对变形字符进行校正处理,通过分析倾斜、错位和弯曲3种变形情况下的D值曲线图,可在字符产生变形时有效判断变形类别,并提出了针对倾斜、错位和弯曲3种变形的处理方法。应用实例表明,该方法适用于背景复杂情况下的旋转或变形的符号,识别准确率达到满意的程度。

A new method of recognizing point-shaped symbols of maps based on distance transformation is presented.Firstly, numeric character is picked up through the distance of key point between two numeric characters.Secondly, three measures is given to dispose the distortion of point-shaped symbols.The application results show that this algorithm can be applied to identify rotated or distortional symbols in the condition of complex background, and the identification accuracy is satisfactory.