基于模糊匹配的手写文字识别平台设计  

Design of Handwritten Character Recognition Platform Based on Pattern Matching

作　　者：杜林 安媛[2] Du Lin;An Yuan(Anhui Jiyuan Testing Technology Co.,Ltd.,Hefei Anhui 230088,China;Xuzhou University of Technology,Xuzhou Jiangsu 221008,China)

机构地区：[1]安徽继远检验检测技术有限公司,安徽合肥230088 [2]徐州工程学院,江苏徐州221008

出　　处：《信息与电脑》2020年第2期101-103,共3页Information & Computer

摘　　要：笔者首先介绍了手写字符识别的背景以及研究方向,重点阐述了图像处理提取特征的方法以及BP神经网络的原理及应用。识别系统分为3个部分,分别是图像的预处理、字符的特征提取以及BP神经网络的训练和识别。其中,预处理主要由对图像进行灰度化、二值化、去噪处理、分割图像并对字符大小归一化这几部分组成。预处理后能够排除大部分对识别的干扰因素,可以降低字符特征提取的难度。本文采用了整体贯穿法来获取图像特征,能更好地提高识别效果。最后用MATLAB GUI来实现BP神经网络的训练与识别,并建立了简单方便的人机交互界面。本系统的训练样本数据为100个各式手写数字图片,测试的数据为100张手写字符图片。经过系统测试发现,对0~9这几个手写数字的识别正确率能达到89%。Firstly,the background and research direction of handwritten character recognition are introduced,especially the feature extraction of image processing The principle and application of BP neural network.The recognition system is divided into three parts:image preprocessing,character feature extraction and BP neural network training and recognition.Among them,preprocessing is mainly composed of graying,binarization,denoising,image segmentation and character size normalization.After preprocessing,most of the interference factors can be eliminated,which can reduce the difficulty of character feature extraction.In this paper,the whole through method is used to obtain image features,which can better improve the recognition effect.Finally,Matlab GUI is used to realize the training and recognition of BP neural network,and a simple and convenient human-computer interface is established.The training sample data of this system is 100 kinds of handwritten digital pictures,and the test data is 100 handwritten character pictures.Through the test of the system,it is found that the recognition accuracy of the handwritten numbers from 0 to 9 can reach 89%.

关 键 词：图像识别 神经网络 手写数字 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]