Research On Application Of Image Matching Based On Local Invariant Features In Recognition Of A New Image-based CAPTCHA

Not long ago, A domestic ticket website introduced a new type of image-based CAPTCHA on its login and booking interface, the difficulty of identify the CAPTCHA is obviously improved and effectively strengthen the security of its website after revised. Since Scale Invariant Feature Transform and Speed-Up Robust Feature have good robustness to blur, angle of view, rotation, scaling, compression, etc. Research on the recognition technology of the new image verification code has important theoretical significance and practical value for promoting the development of the technology of CAPTCHA and enhancing the security of Internet. We try to introduce these two algorithms to the application of new image-based CAPTCHA recognition and some experiments are carried out, expected to obtain a good recognition. The work and contributions of this paper are as follows:Firstly, this paper mainly studies the principle of two well-known invariant features, namely SIFT and SURF feature detection, and we analyze the interference factors in the new image verification code, including blur, angle of view, rotation scaling, compression, etc. And 5 sets of experiments were designed to test the performance of the algorithms, and the advantages and disadvantages of various transformation scenarios in the new image-based CAPTCHA.Then, this paper designs a new image-based CAPTCHA system, and uses the image matching technology based on SIFT and SURF to accomplishment the recognition of the system. The experimental are compared with 50 pixels level and 100 pixels level for 6 kinds of image verification code with different object types. The results show that the algorithms can effectively identify the image-based CAPTCHA which contains a certain area of the object type. And the performance of these two kinds of feature detection algorithms for the identification of this new image-based CAPTCHA is largely determined by the object type, resolution and interference of background, etc.