Design and Implementation of Second-generation ID Card Number Identification Model based on TensorFlow

摘要

This paper designs and implements a model base on tensorflow framework in Python language to complete the identification of ID card number. Firstly, the image of ID card is preprocessed by a computer vision library (OpenCV), including angle correction, grayscale and binarization and so on. Secondly, an algorithm is used to project the image, and obvious projection peak area is number area, meanwhile the number area of ID card is determined, character segmentation is carried out by using the fixed width height ratio and spacing of characters. And then the data set is trained by the convolutional neural network. Finally, the character segmentation is recognized and the ID card number is output. In the end, the model is tested to show that the training speed is fast, the recognition and identification accuracy is high, and the convolutional neural network(CNN) has a certain value.