基于Hopfield神经网络模型的验证码识别问题

Hopfield Neural Network Based Approach to Recognizing the Verification Code

摘要: 为了避免恶意破解密码、刷票、论坛注水、黑客攻击等行为，验证码因此而生，它通过强制人机交互的方法来抵御机器自动化攻击。但准确识别验证码是当前还未完全解决的难题。本文的研究对象为四位数字的验证码，采用分割分析和Hopfield神经网络的方法对验证码识别及其精度进行分析。建立并训练Hopfield网络对验证码图像变换后的矢量进行模式识别，比对的结论表明，Hopfield神经网络算法可以得到比较好的测试结果，具有推广应用的价值。

Abstract: In order to avoid malicious cracking of passwords, ticket brushing, forum water injection, hacking, etc., verification codes are born. It uses the method of forcing human-computer interaction to resist machine automation attacks. But accurately identifying the verification code is a difficult problem that has not yet been completely resolved. The research object of this article is a four-digit verification code. The method of segmentation analysis and Hopfield neural network is used to analyze the verification code recognition and its accuracy. Establish and train the Hopfield network to perform pattern recognition on the transformed vector of the captcha image. The conclusion of the comparison shows that the Hopfield neural network algorithm can get better test results and has the value of popularization and application.