特征信息的新数字图像识别方法  

New Algorithm of Numeral Image Recognition Based on Feature Information

作　　者：马斌[1] 边树海[1] 王长涛[1] 韩忠华[1] 孟庆斌[1] 

机构地区：[1]沈阳建筑大学信息与控制工程学院,辽宁沈阳110168

出　　处：《沈阳建筑大学学报（自然科学版）》2011年第6期1187-1192,共6页Journal of Shenyang Jianzhu University：Natural Science

基　　金：国家十一五科技支撑计划项目(2008BAJ08B08-3);住房和城乡建设部科技项目(2010-K9-49)

摘　　要：目的为提高数字图像的总体识别率,对数字图像识别方法进行了深入研究.方法在对原始数字图像进行保守平滑、均值滤波、高斯锐化和二值化等处理的基础上,通过提取数字与中垂列的交点数、与1/3和5/6字高处的水平穿线的交点数、与将数字上部分和下部分均k等分的2*(k-1)条水平穿线交点数等特征信息.结果该特征比较的数字图像识别方法对0～9这10个数字的总体识别率可以达到96%以上.结论特征比较的数字图像识别方法简单、识别速度快、容易实现且识别效果好.In order to improve the recognition rate of numeral image,the paper deeply studied numeral image recognition method.The characteristics information of the number were extracted such as the intersection number of figure and its middle vertical column,the intersection numbers of figure and level threads of 1/3 and 5/6 figure height,and the intersection numbers of figure and its 2＊（k-1）level threads which make up and down figures k equal sections,after the process of conservative smoothing,mean filtering,Gaussian sharpen and binarization to the original numeral image.The experimental results show that recognition rate of the new approach is up to 96%.This method is simple,fast-speed,easy to implement and has good recognition effect.

关 键 词：数字图像识别 特征信息 识别率 特征比较 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]