基于BP神经网络手写体数字识别的软件设计
摘 要
手写体数字识别是利用机器或计算机自动辨认手写体阿拉伯数字的一种技术,是光学字符识别技术的一个分支。由于阿拉伯数字的世界通用性,并且数字的识别和处理也常常是一些自动化系统的核心和关键,所以对手写体数字识别研究通用性强,且意义重大。
本文主要的研究的工作集中在图像预处理和选择合适的特征向量,并实现一个完整手写体数字识别系统。本文中对几种常见的二值化算法进行比较,并最终选择基于梯度的二值化算法;在本文中,提出了一种方法来解决结构点检测的传统方法的缺陷。另外本文还提出将一般用来直接识别字符的凸凹特征作为字符的特征向量之一。
通过对NIST的数据进行测试,实验数据表明本文设计的数字识别系统对手写体数字识别具有较高的识别率。
关键词:手写体数字识别,特征向量,二值化算法,凸凹特征
The Software Design of Handwritten Numeral Recognition
Abstract
The handwritten numeral recognition is a technology, which auto recognizes the handwriting Arabian numeral via machines or computers, and a special field in the Optical Character Recognition technology. Then handwritten numeral recognition research is greatly general-purpose and significative, because of the universal Arabic numerals. On the same score, the handwritten numeral, recognition technologies are playing an important role in a number of automatization systems.
In this paper, the main study focused on image pre-processing and selection of appropriate feature vectors, and to realize a complete system of handwritten numeral recognition. There are several common comparisons of binarization algorithm in this article, and choose the gradient-based binarization algorithm finally; and objecting to the defection of traditional methods of structural point detection, a solution to the problem was put up in this article. In addition, this paper also raises the convex-concave feature as one of the character feature vectors.
Through the NIST test data, experimental data shows that the digital identification system designed for handwritten numeral recognition has a high recognition rate.
Keywords: Handwritten numeral recognition, Character feature vectors, Binarization algorithm, Convex-concave feature