基于SVM的手写数字识别的应用与实现【毕业论文】.docVIP

本科毕业论文 （20 届） 基于SVM的手写数字识别的应用与实现 专业：计算机科学与技术 摘要 手写数字识别是字符识别的一个分支，虽然只是识别简单的10个数字，但却有着非常大的实用价值。在文献检索、邮政系统、办公自动化、表格录入、银行票据处理等方面都有着广泛的应用。手写数字虽然只有10个种类，但很多情况下，对识别的精度要求非常高，而且每个人都有不同的字迹，想要做到准确的识别，就很有难度了。 支持向量机(Support Vector Machine，SVM)最初于20世纪90年代由Vapnik提出，是基于统计学习理论的一项新技术，是借助于最优化方法解决机器学习问题的新工具，具有许多引人注目的优点和有前途的实验性能，在人脸识别，手写数字识别，网页分类等方面有非常理想的效果。所以，本文将支持向量机应用于手写数字识别，希望能够得到不错的精度。 手写数字识别一共有10类样本，是一个多分类问题，本文将多个支持向量机的二类分类器组合起来构成一个多类分类器，以便实现手写数字识别。我们把样本分为训练样本集和测试样本集，训练样本集与测试样本集都要经过预处理得到点阵数据，然后再经过特征提取，用来训练识别器以及进行识别。 【关键词】手写数字识别，支持向量机，分类器，特征提取，预处理 Application and Implementation of Handwriting Numerals Recognition Based on SVM Abstract Handwriting numerals recognition is a branch of the character recognition, although just recognition simple 10 figures, but it has a very big practical value. In the literature retrieval, the postal system, office automation, form entry, bank bills are processing, etc in a wide range of applications. Handwriting numerals although only 10 species, but in many cases, the accuracy requirement of recognition is very high, and everybody has different handwriting, to get accurate identification, is very difficult. The Support Vector Machine (SVM) in the early 1990s, is put proposed by Vapnik, in data mining is a new technology, is a further optimization method of machine learning problem solving new tools, has many compelling advantages and promising experimental performance, in face recognition, Handwriting numerals identification, classification of web pages is very ideal effect. So, this paper will support vector machine applied in Handwriting numeral recognition, the hope can obtain good accuracy. Handwriting numerals identifying a total of ten kinds of samples, is a multiple classification problem, this paper will more support vector machine classifier ii combined form a multitude of classifier, in order to realize Handwriting numeral recognition. We put the samples are divided into the training sample set and testing samples, the training sample se