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.%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.
展开▼