基于图像识别和神经网络的验证码识别

来源期刊:中南大学学报(自然科学版)2011年第z1期

论文作者:连晓岩 邓方

文章页码:48 - 52

关键词:验证码;图像识别;神经网络;预处理;字符;分割;识别

Key words:CAPTCHA; image recognition; neural networks; pre-operating; character; segmentation; recognition

摘    要:提出一种基于图像识别和神经网络的验证码识别方法,能快速、高效地识别兼有数字、英文字母等多种字符的验证码图像。首先,对验证码图像进行灰度化、二值化和去噪等预处理操作,滤除干扰信息、突出字符特征,得到含较少噪声的二值图像;其次,采用一种经过改进的、融合投影特性的连通域分割法获得单个字符图像并标准化;再次,对标准字符图像进行特征提取,构建并训练神经网络;最后,测试神经网络训练效果实现字符识别,在Matlab环境下处理不同网站的大量验证码图像。实验结果表明:本文提出的方法对于字符排列不规则、干扰噪声较多的验证码具有很高的识别率和有效性。

Abstract: Based on image recognition and neural networks (NNs), a rapid and effective CAPTCHA recognition method was proposed which has different characters of numbers and English letters. Firstly, pre-operating was put in practice, including graying, binaryzation and removing noise to get the binary image with higher quality. Secondly, images with single character were acquired through an improved segmentation algorithm combining connected domain and projection together, and then normalized. Thirdly, features of character image samples were extracted; the NNs were built and trained. Finally, the trained NNs were tested to realize character recognition. Based on the software environment of Matlab, the recognition processing and result of many CAPTCHA images from different websites were proposed. The experiments show that the method is effective and feasible for the CAPTCHAs which contain irregular placed characters and much noise.

有色金属在线官网  |   会议  |   在线投稿  |   购买纸书  |   科技图书馆

中南大学出版社 技术支持 版权声明   电话:0731-88830515 88830516   传真:0731-88710482   Email:administrator@cnnmol.com

互联网出版许可证:(署)网出证(京)字第342号   京ICP备17050991号-6      京公网安备11010802042557号