Identification Method Of ID Card Information

Research on identification method of ID Card information.First,the user login module of the XinJiang horse industry innovation platform requires real-name authentication.Real-name authentication is to automatically identify the identity information of the registered user by uploading the id photo.Users who have the real-name authentication can have more functional rights on the platform.Second,there are differences in the shooting scenes,angles,equipment and photographers of id photos,so traditional id information identification methods can no longer meet the requirements of identification accuracy of id information in complex environments.In order to further improve the accuracy of ID Card identification,to overcome the external environment interference of ID Card information identification.In this paper,based on the method of deep learning ID Card information recognition,the corresponding neural network model is trained by deep learning algorithm to detect and recognize the ID Card information.First of all,in order to prevent the real-name authentication upload images in the process of the ID Card,but only id information of images,text before to test the id information recognition,it is necessary to identify pictures and judge whether the image for the identification of qualified photo text image,and not just the information based on the deep learning framework using ResNet residual neural network model to identify id photos,classifying input computer pictures for ID Card,if the judgment for the ID Card to the next step photo id information detection operation.Through the comparison experiment of ResNet residual network resnet-50,resnet-101 and resnet-152 at three different levels,the identification of the ID Card photo of resnet-101 model was finally adopted.The identification accuracy of the ID Card can reach 92.00%according to the experiment.Secondly,in this paper,using the method of deep learning training Text detection under complex scene CTPN neural Network model(Text Proposal Network)to test the id information,detect the Text area.Especially due to the influence of illumination detection of ID Card,in order to improve the model of influence factors of light id information detection of adaptation,this experiment in view of the partial data to enhance data processing,mainly on ID Card image color,contrast,saturation,chroma processing,data set by Labellmg label production,using VGG-16(Visual Geometry Group Network Visual Geometry Group Network)to extract the id space information,RNN(Recurrent Neural Networks)is used to extract sequence information,and the information location is determined by text line construction algorithm.Text information can be well detected by tests on the test data set.Target detection evaluation index AP is used to evaluate the detection model,and the AP value of information detection evaluation index reaches 70.34%.At last,DenseNet(Dense Connected Convolutional Networks)was trained torecognize the detected id information through the deep learning method.According to the characteristics of the ID Card information,all kinds of information on the identity Cards for variable length characters,so character recognition method is used to complete information identification task,using a large number of variable length data sets of character image model of training,finally on the test data set recognition accuracy rate reached 98.50%,due to the training data set is grayscale images,the model of gray level images after processing more resilient,and further improve recognition effect.