基于启发式GA-SVM的手写数字字符识别的研究  

Research on Handwritten Numeral Character Recognition Based on Heuristic GA-SVM

作　　者：石会芳[1] 胡小兵[1] 刘瑞杰[1] 叶剑英[1] 

机构地区：[1]重庆大学数学与统计学院,重庆401331

出　　处：《计算机技术与发展》2012年第10期5-9,共5页Computer Technology and Development

基　　金：重庆市自然科学基金资助项目(CSPC;2005BB2197);重庆大学"211工程"三期创新人才培养计划建设基金资助项目(S-09110)

摘　　要：数字字符识别技术是图像处理领域中的一个重要研究方向,并且在社会经济生活的许多方面有着越来越广泛的应用。预处理是手写数字识别中重要的一个环节,它的优劣直接关系到识别算法的性能。文中提出了一种改进后的数字字符识别处理方案。该方案首先对输入的数字字符图像进行预处理,然后使用大津法进行全局图像阈值选取,通过阈值处理将灰度图像转化为二值图像。最后利用libsvm加强工具箱和启发式遗传算法进行手写数字的识别。实验结果显示该方法提高了识别数字的准确度。The technology of numeral character recognition is an important research direction in image processing area and the application in various fields is more and more widely. Pretreatment is a vital step in recognition of handwritten numerals, its strengths and weaknes- ses is directly related to the performance of the recognition algorithm. It describes an improved treating plan of numeral character recogni- tion. Firstly, the method is to do the pretreatment for the input numeral character image, and then Otsu method is used to finish the glob- al image threshold selection, through threshold processing shift grayscale image into binary image. Finally, use libsvm enhance toolbox and the method of genetic algorithm to discriminate digit. Experimental results show that the accuracy of the numeral recognition is improved.

关 键 词：数字字符识别 阈值处理 支持向量机 遗传算法 

分 类 号：TP31[自动化与计算机技术—计算机软件与理论]