The Research On Recognition Technology Of CAPTCHA Based Character Distortion And Adhesion

CAPTCHA is the standard technology to tell computer and human apart in the fieldof security, The most common scheme is based on the text of letters and numbers. Theprinciple of connecting adjacent characters provides a high security for the CATPCHA.The typical scheme is Yahoo！, Baidu and reCAPTCHA. It has a good reliability to resistthe character segmentation and recognition. In this paper, we identifications sometypical schemes by utilizing the characteristics of these CAPTCHAs. Early studiesbreak the CAPTCHA by separating the original string into pieces one by one andrecognizing them using the technology of OCR. In this paper, the recognized characterwill be removed from the original string step by step. The preprocessing, projectiontechnology, closed-loop detecting, four-line spectrum and curve fitting is used in therecognition process. The success split rate of Yahoo CAPTCHA is78%and therecognition rate is54.7%. The recognition rate of Baidu CAPTCHA is34%and therecognition rate of reCAPTCHA is17%. As the ideal CAPTCHA recognition rateshould be less than0.01%by the computer, the segmentation and recognition in thispaper is successful.