基于Faster-RCNN的极验点选式验证码识别

doi:10.16180/j.cnki.issn1007-7820.2019.09.009

电子科技 ›› 2019, Vol. 32 ›› Issue (9): 42-46.doi: 10.16180/j.cnki.issn1007-7820.2019.09.009

基于Faster-RCNN的极验点选式验证码识别

周文凯,韩芳,孔维健

东华大学信息科学与技术学院,上海 201620

收稿日期:2018-09-12 出版日期:2019-09-15 发布日期:2019-09-19
作者简介:周文凯(1994-),男,硕士研究生。研究方向:图像处理,虚拟现实。|韩芳(1981-),女,博士,教授。研究方向:智能系统与神经动力学。|孔维健(1983-),男,博士,讲师。研究方向:计算智能,复杂工业过程建模与优化。
基金资助:
国家自然科学基金(11572084,11472061)(11572084);国家自然科学基金(11572084,11472061)(11472061);中央高校基本科研业务费(18D210402)

Point-selective Geetest CAPTCHA Recognition Based on Faster-RCNN

ZHOU Wenkai,HAN Fang,KONG Weijian

School of Information Science and Technology,Donghua University,Shanghai 201620,China

Received:2018-09-12 Online:2019-09-15 Published:2019-09-19
Supported by:
National Natural Science Foundation of China(11572084);National Natural Science Foundation of China(11472061);The Fundamental Research Funds for the Central Universities(18D210402)

摘要/Abstract

摘要：

针对传统方法难以识别极验点选式验证码的问题,文中提出一种基于Faster-RCNN目标检测模型和卷积神经网络的识别方法。通过简化的RPN提高Faster-RCNN对于背景图片文本定位的精度,再设计卷积神经网络对文本进行分类识别,并训练Tesseract-OCR识别库对信息提示文本进行识别,实现背景图片文本识别结果与信息提示文本识别结果一一对应,达到识别此类验证码的目的。实验结果表明,该方法识别此类验证码的准确率达到72.4%。

关键词: 极验验证码, Faster-RCNN, 卷积神经网络, Tesseract-OCR

Abstract:

In view of the problem that traditional methods were difficult to recognize the geetest point-selective CAPTCHA , this paper proposed a recognition method that based on Faster-RCNN object detection model and convolutional neural network. Firstly, it improved the accuracy of text positioning by simplified RPN, then designed convolutional neural network to classify and identify texts,and trained the Tesseract-OCR identification databases to recognize the message texts. Finally, the result of texts recognition which embedded into background picture corresponded to the result of message texts recognition, then it reached the purpose of CAPTCHA recognition. Experimental results showed that the accuracy of this recognition method up to 72.4%.

Key words: geetest CAPTCHA recognition, Faster-RCNN, convolutional neural network, Tesseract-OCR

中图分类号:

TP391

周文凯,韩芳,孔维健. 基于Faster-RCNN的极验点选式验证码识别[J]. 电子科技, 2019, 32(9): 42-46.

ZHOU Wenkai,HAN Fang,KONG Weijian. Point-selective Geetest CAPTCHA Recognition Based on Faster-RCNN[J]. Electronic Science and Technology, 2019, 32(9): 42-46.

图/表 10

图1

图2

图3

图4

图 5

表1

表2

表3

图6

图7

参考文献 16

[1]	胡健, 柳青, 王海林 . 验证码安全与验证码绕过技术[J]. 计算机应用, 2016,36(S1):37-41
	Hu Jian, Liu Qing, Wang Hailin . CAPTCHA security and bypass technique[J]. Journal of Computer Applications, 2016,36(S1):37-41.
[2]	孟凯 . 基于SVM的空心验证码识别技术研究[D]. 重庆:重庆邮电大学, 2017.
	Meng Kai . Research on recognition technology of hollow CHPTCHAs based on SVM[D]. Chongqing: Chongqing University of Posts and Telecommunication, 2017.
[3]	王斌君, 王靖亚, 杜凯选 . 验证码技术的攻防对策研究[J]. 计算机应用研究, 2013,30(9):2777-2779.
	Wang Binjun, Wang Jingya, Du Kaixuan . Research on attach strategy of CHPTCHA technology[J]. Application Research of Computers, 2013,30(9):2777-2779.
[4]	汪中 . 验证码识别技术及应用[J]. 计算机光盘软件与应用, 2014,17(5):167-168.
	Wang Zhong . CHPTCHA recognition technology and application[J]. Computer CD Software and Application, 2014,17(5):167-168.
[5]	尹龙 . 扭曲粘连字符验证码识别研究[D]. 合肥:中国科学技术大学, 2014.
	Yin Long . Recognition of distorted and merged text-based CHPTCHA[D]. Hefei:University of Science and Technology of China, 2014.
[6]	安改换 . 扭曲粘连字符验证码识别技术研究[D]. 上海:上海应用技术大学, 2016.
	An Gaihuan . Recognition technology research of distorted and merged text-based CHPTCHA[D]. Shanghai:Shanghai Institution of Technology, 2016.
[7]	唐海涛 . 自组织增量神经网络的验证码识别模型与算法[D]. 广州:广东工业大学, 2016.
	Lu P, Shan L, Li J , et al. A new segmentation method for connected characters in CAPTCHA[C].Changshu:International Conference on Control, Automation and Information Sciences ( ICCAIS), 2015.
[9]	汪洋, 许映秋, 彭艳兵 . 基于KNN技术的校内网验证码识别[J].计算机与现代化, 2017(2):93-97.
	Wang Yang, Xu Yingqiu, Peng Yanbing . KNN-based verification code recognition on campus network[J]. Computer and Modernization, 2017(2):93-97.
[10]	Wang Ye, Lu Mi. A self-adaptive algorithm to defeat text-based CAPTCHA [C].Taipei:IEEE International Confer-ence on Industrial Technology, 2016.
[11]	Garg G, Pollett C . Neural network CAPTCHA crackers[C].San Francisco:Future Technologies Conference (FTC), 2016.
[12]	刘达荣, 张远平, 汤茂斌 , 等. 基于渐进式学习的神经网络端到端验证码识别[J].计算机技术与发展, 2018(9):9-13.
	Liu Darong, Zhang Yuanping, Tang Maobin , et al. End-to-end verification code identification of neural network based on progressive learning[J].Computer Technology and Development, 2018(9):9-13.
[13]	熊海朋, 陈洋洋, 陈春玮 . 基于卷积神经网络的场景图像文本定位研究[J]. 电子科技, 2018,31(1):50-53.
	Xiong Haipeng, Chen Yangyang, Chen Chunyu . Text location in image based on convolutional neural network[J]. Electronic Science and Technology, 2018,31(1):50-53.
[14]	Ren S, Girshick R . Faster R-CNN:towards real-time objection with region proposal networks[J]. IEEE Transactions on Pattern Analysis & Machine Intelligence, 2017,39(6):1137-1149.
[15]	Simonyan K, Zisserman A. Very deep convolutional networks for large-scale image recognition [C].San Diego:International Conference on Learning Representations, 2015.
[16]	Li Q, An W, Zhou A , et al. Recognition of offline handwritten Chinese characters using the tesseract open source OCR engine[C].Hangzhou:8 ^th International Conference on Intelligent Human-Machine Systems and Cybernetics (IHMSC), 2016.

特征提取网络	RPN	误识率	时间/ms
ZF网络	使用原始的RPN	25%	105
	使用改进的RPN	17%	94
VGG网络	使用原始的RPN	14.3%	198
	使用改进的RPN	9.4%	192

类型	层数	卷积核数量	卷积核大小	卷积核步长	输出大小
Conv1	2	64	3×3	1	64×64
Pool_1	1	64	3×3	2	32×32
Conv2	2	128	3×3	1	32×32
Pool_2	1	128	3×3	2	16×16
Conv3	3	256	3×3	1	16×16
Pool_3	1	256	3×3	2	8×8
Fc	1	-	-	-	4 096
Softmax	1	-	-	-	2 059

方法	识别率
不使用数据增广	92.25%
不使用批标准化	87.98%
唯一的预测值	95.66%
本文方法	97.94%

[1]	李辉,王一丞. 基于CNNCIFG-Attention模型的文本情感分类[J]. 电子科技, 2022, 35(2): 46-51.
[2]	程晓雅,张雷. 基于深度学习的有遮挡人脸识别方法研究[J]. 电子科技, 2022, 35(1): 35-39.
[3]	战荫伟,朱百万,杨卓. 一种车脸识别算法的研究与应用[J]. 电子科技, 2021, 34(8): 1-7.
[4]	廖璐明,张伟. 基于改进VGG16网络的混合批量训练交通标志识别[J]. 电子科技, 2021, 34(8): 8-13.
[5]	马立新,豆晨飞,宋晨灿,杨天笑. 基于特征融合卷积神经网络的绝缘子无损检测[J]. 电子科技, 2021, 34(7): 26-30.
[6]	顾伟,李菲菲,陈虬. 基于多特征融合的行人检测方法[J]. 电子科技, 2021, 34(5): 29-34.
[7]	薛永杰,巨志勇. 基于改进AlexNet的鱼类识别算法[J]. 电子科技, 2021, 34(4): 12-17.
[8]	闫书豪,乔美英. 基于一维WConv-BiLSTM的轴承故障诊断算法[J]. 电子科技, 2021, 34(4): 75-82.
[9]	葛靖,刘子龙. 基于CNN和LSTM的睡眠呼吸暂停检测算法[J]. 电子科技, 2021, 34(2): 21-26.
[10]	朱斌,刘子龙. 基于新型初始模块的卷积神经网络图像分类方法[J]. 电子科技, 2021, 34(2): 52-56.
[11]	张灿,陈玮,尹钟. 基于弱监督宫颈细胞图像的语义分割方法[J]. 电子科技, 2021, 34(12): 68-74.
[12]	张莹,刘子龙,万伟. 基于Faster R-CNN的无人机车辆目标检测[J]. 电子科技, 2021, 34(11): 11-20.
[13]	李诚,刘昊,蒋希峰,吴军法,韩文刚,高建国. 基于VGG网络的发电机定转子智能诊断算法[J]. 电子科技, 2021, 34(11): 62-66.
[14]	董波,周燕,王永雄. 基于渐进结构感受野和全局注意力的显著性检测[J]. 电子科技, 2021, 34(1): 23-30.
[15]	张翔,史志才,陈良. 基于SWA优化级联网络的表情识别方法[J]. 电子科技, 2020, 33(9): 16-20.

基于Faster-RCNN的极验点选式验证码识别

Point-selective Geetest CAPTCHA Recognition Based on Faster-RCNN

RichHTML

PDF (PC)

赞

可视化

摘要/Abstract

引用本文

使用本文

图/表 10

参考文献 16

相关文章 15

Metrics

本文评价

推荐阅读 10