摘要
图像中手写数字的识别一直被人研究探索,目前国内外提出的识别方法大致分为两种:基于BP神经网络统计特征识别算法、基于几何形体结构特征识别算法。文章在前人探索的形体结构特征算法基础上进行改进优化,利用孔洞、端点、交叉点等特征进行识别外,同时发现新的特征、新的方法:离心率、行列扫描区域数量、区域骨架搜索。组合这些形体特征和方法,可完成手写数字的识别。实验证明手写数字识别在识别准确率上优于先前传统形体结构特征方法。
Handwritten recognition in digital images has been studied at home and abroad, recognition methods proposed can be divided into two types： one is statistical feature recognition algorithm based on BP neural network, the other is structural feature recognition algorithm based on geometric structure. This paper improves and optimizes the algorithm based on the structure and feature of the previous researchers, takes advantage of holes, endpoint, cross point feature and discover new features and new method, such as eccentricity, the number of row scanning region and region skeleton search. The combination of these form characteristics and methods is able to complete handwritten numeral recognition. The experimental results show that the recognition accuracy of handwritten numeral recognition is better than that of the traditional form structure feature method.
出处
《无线互联科技》
2017年第11期92-94,共3页
Wireless Internet Technology
基金
郑州大学大学生创新创业训练计划资助项目(国家级项目)
项目名称:基于机器视觉的试卷统分登分及成绩分析系统
项目编号:201610459047
关键词
手写数字识别
形体结构特征
离心率
行列扫描区域数量
区域骨架搜索
handwritten numeral recognition
form structural feature
eccentricity
number of row scanning region
region skeleton search
