摘要
对基于结构特征的数字识别方法进行了改进,首先根据投影法准确定位出数字区域,接着对图像进行预处理、二值化、区域定位及字符分割等操作,最后根据数字结构特征对字符进行分类和识别.该方法不需要对图像进行倾斜校正、亮度调整、去噪等操作,节省了大量时间.实验结果表明,该算法准确率达到98%,耗时仅为5.52ms,满足实时性要求.
Digit recognition method based on structural features is improved.First,according to the projective method to locate the number zone accurately.Then it preprocesses the image,including binarization,zone location,character segmentation and so on.This method does not require brightness adjustment,skew correction,denoising,saving much time.The experimental results show that the recognition rate can reach 98%,while the recognition time is only 5.52 ms.
出处
《云南师范大学学报（自然科学版）》
2017年第1期37-42,共6页
Journal of Yunnan Normal University:Natural Sciences Edition
基金
云南省科技惠民计划资助项目(2014RA042)
关键词
快递单
结构特征
数字分割
数字识别
Express lists
Structural features
Number segmentation
Number recognition
