The Development And Implementation Of Identification System On Mobile Intelligent Terminal

The traditional Information Input requires someone to record information,and then the operator enter the vital inform ation into the com puter.The another way to input information isto go to the design ated place and upload th e scanned d ocuments.The former way not restrict the place of infor mation entry,but cost a lar ge amount of manpower and material resources.This way is error-prone and inefficient.Although the latter way has higher ef ficiency and accuracy,but the site is relatively fixed.In the current social situation,most of the documents' information are stored in the cards.The urgent needs of all walks of life are identifying information rapidly and expediently.The emergence of s martphone makes this fancy possible.The inform ation identification system of the smartphone can be widely used in service industry,transportation system,public security system.The information identification system can save a lot of manpower and resources.This system also improves the efficiency and accuracy of the information and has widespread application prospects.How to make a good extraction and recognition of the text information in different documents is the vital problem of this identification system.For identifying a document image,the first task is to extract the key information.In this paper,we design a different image pre-processing method for dif ferent documents to ensure that the document information can be ex tracted correctly.In this paper,we usethe binarization algorithm based on the approximation of character stro ke width.Thisalgorithm can ef fectively reduce the impact of background,noise,reflec tionthe,and improve the recognition rate.According to the characteristics of different characters in the information recognition,we use two kinds of popular m ethods to iden tify the text inform ation.We use the Tesseract-OCR engine to identify the English and numbers.This characters have simple structure and fewer categories.The accuracy of using the Tesseract-OCR engine has met the requirements and the size of character set is small,we can use this set in sm artphone expediently.But Chinese characters have complicated structure and more categories.The recognition rate of using Tesseract-OCR engine is not higher enough and the size of the character set is large.This paper presents a Chinese character r ecognition method based on feature extraction and convolutional neural network which com bines the traditional feature extraction method with neural network.This method makes up the loss of feature information in the process of using neural network training.Then,we use the Dropout technology in the input laye r and the hidden layer to pr event the network from over fitting and improve the generalization ability of the network.The recognition accuracy of this method is higher than using T esseract-OCR engine and the size of character set is also smaller.It is easy to be transplanted to the smartphone.In this paper,an identification system is developed based on the above requirements.At present,it supports the id entification of positive and ne gative identification,driving license recognition.This system is divided into the Android version and the iOS version.And it su pports the vast majority of smart devices on the m arket.The system can successfully identify documents in English,numbers,Chinese.English recognition rate reached about 98.4%.The recognition rate of num bers reached abo ut 99.2%.The Chinese recognition rate reach ed about 98.27 %.And the overall id entification rate of documents is about 90%.