Research On The Generating Technology Of Captcha Based On Image Style Transformation

The sky image of Van Gogh and a picture of a cute cat,but a cat picture with the style of Van Gogh's sky seems quite interesting too.Under the thought of image style transformation,except the general purpose of “adding a filter to a photo”,it can also be applied in many other ways.Based on this idea,this paper proposed a generating technology of image captcha.Due to the continuous developing society,the captcha is more and more important in our daily life.The captcha can protect the personal information effectively through distinguishing whether the user is a machine or a person.For example,it can reduce the risk of the website being attacked,the flooding phenomenon of the forum website and etc.So a good generating technology of captcha should have two features:(1)friendly to human;(2)unfriendly to the machine.The generating technology described in this paper can satisfy the features above,and it is based on the VGG network model and Transfer Learning.The work of this paper is divided into three parts: 1.Deep convolutional neural network can be used in image recognition,inspired by it and combined with the VGG network model,this paper proposes a method in style transformation of an image,and optimize the value of parameters through lots of experiments.2.Generating image captcha under the premise of the algorithm of image style transformation,and continuously improve the algorithm combined with the reality of image captcha.3.Compared to the captcha which is used in reality,it can prove that it is useful of the method proposed in this paper.