Captcha automatic segmentation and recognition based on improved vertical projection

• Published in: 2017 IEEE 9th International Conference on Communication Software and Networks (ICCSN) pp. 1167 - 1172
• Main Authors: Lili Zhang, Yuxiang Xie, Xidao Luan, Jingmeng He
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01-05-2017
• Subjects: Algorithm design and analysis; captcha; CAPTCHAs; Character recognition; conglutinate characters; Histograms; Image edge detection; Image segmentation; Interference; recognition; segmentation
• ISSN: 2472-8489
• DOI: 10.1109/ICCSN.2017.8230294

Captcha recognition plays an important role in Internet security, and conglutination characters segmentation is still the bottleneck of captcha recognition research. In this paper, a method of captcha segmentation based on improved vertical projection can efficiently solve the segmentation problem of different types of conglutination characters in captcha combining both numbers and letters, so as to improve the accuracy of captcha recognition. Firstly, take preprocessing to captcha image including removing interfering background, denoising and binarization, getting binary captcha image with less noise. Secondly, apply improved vertical projection method to captcha characters segmentation, propose targeted segmentation method to different conglutinate type characters, getting split single character image. Thirdly, take the overlap rate of sample and template character pixels as matching rate, using template matching algorithm for recognition. Finally, through program experiment to achieve the algorithms of segmentation and recognition proposed above. Experimental result suggests that the method proposed in this paper has efficient segmentation effect on conglutinate characters, consequently improving the accuracy of captcha recognition.