Deep extension technology: intelligent OCR recognition technology based on deep learning has great potential

intelligence OCR Technology is one of the important atomic capabilities in the field of artificial intelligence , Play a role in the process of industry transformation . In recent years , Mobile Internet 、 New technologies such as big data are developing rapidly , It forces traditional industries to move towards intelligence 、 Transformation in the direction of mobility . This general trend also promotes tradition OCR Technology continues to move towards intelligence OCR upgrade , To meet more complex 、 More demanding task requirements .

With the integration of deep learning , Realized OCR Recognition of complex text , That is intelligence OCR distinguish . At the same time, it shall cooperate with the comprehensive application of other information means , Make intelligent OCR With mobile end adaptation 、 Multitask detection 、 Whole line identification 、 Image segmentation, positioning and classification , More extensive application scenarios , In the field of Library and information, book text and card identification will be more efficient .

01 Tradition OCR Identify emerging technical defects

Optical character recognition （Optical Character Recognition,OCR） It refers to the automatic recognition of the text content in the image , It belongs to an important branch of artificial intelligence machine vision , That is, the text 、 The characters on cards and certificates are transformed into electronic data recognized by computers through optical and other technical means . Tradition OCR Identify statistical patterns , The processing flow is long , Including image preprocessing 、 Two valued 、 Connected domain analysis 、 Layout analysis 、 Line segmentation 、 Word segmentation 、 Single character recognition and post-processing . Typical tradition OCR The identification process is shown in the figure 1 Shown .

chart 1 Tradition OCR Identify technical processes

Tradition OCR The identification method has many disadvantages , The summary is as follows ：a） During layout analysis , Use a lot of rules , This leads to high program maintenance costs .b） Industry domain analysis completely depends on the binary graph obtained by image binarization , The effect of scanning documents is acceptable , In the face of mobile phone shooting and high-speed camera drawing , It is difficult to get a good binary graph , A lot of information is lost in the process of binarization .c） Tradition OCR Technology includes 8 A module , Pictured 1 Shown , The imperfection of any one of these modules will cause errors , The accumulation of errors will lead to a sharp decline in the recognition rate .d） Tradition OCR Poor recognition flexibility , Complex samples taken in natural scenes can hardly be processed , The lifting space has not been modified , Low availability .e） The traditional method will OCR The system is fragmented into too many links , Rely on manual rules , It is necessary to introduce manual intervention in each link and set method parameters according to the scenario , It is difficult to do end-to-end training .

The deep learning algorithm can effectively avoid the traditional OCR Identified deficiencies , A more abstract high-level representation of attribute categories or features by combining low-level features , Distributed feature representation of mining data . Using neural network to simulate human brain for analysis 、 Learning and training , That is to imitate the mechanism of the human brain to analyze images 、 Data such as voice and text , It is widely used in modeling and processing of artificial intelligence .

02 Intelligent learning based on deep learning OCR "

Intelligent learning based on deep learning OCR Technology is a leap forward upgrade , The deep learning algorithm realizes the whole line recognition , Promoted OCR Recognition rate and recognition speed , Text that takes a few minutes to enter manually , intelligence OCR Technology can accurately identify at the speed of seconds . intelligence OCR Identification technology optimizes the identification process , The optimized identification process includes detection 、 Identification and post-processing 3 Main steps , Pictured 2 Shown .

chart 2 intelligence OCR Identify technical processes

Deep extension technology intelligent one-stop OCR The service platform Based on deep learning OCR Location and identification , Through convolution neural network CNN、 Cyclic neural network RNN、 Long and short term memory network LSTM Technical realization , It can automatically locate the text area and recognize the whole line of text on the gray image , Solved the tradition OCR In word recognition technology, it is impossible to judge similar words with the help of context .

Besides , intelligence OCR Recognition technology has significantly improved the tolerance and recognition accuracy of low-quality pictures , Low resolution and fuzzy character recognition in print 、 Print complex or non-uniform background recognition 、 Printed multilingual hybrid recognition 、 Print art font recognition 、 Handwritten lower case digit recognition 、 Handwritten amount recognition in words 、 Realize efficient recognition and classification in scenarios such as handwritten general text recognition .

Deep extension technology intelligent one-stop OCR The service platform is based on the intelligence of deep learning OCR Recognition technology supports image recognition taken by mobile devices , It can be used for focusing inaccuracy 、 High noise 、 low resolution 、 Strong light and shadow and other complex backgrounds .

In addition to the Card identification 、 Note identification 、 Form recognition 、 Document recognition , intelligence OCR It can be applied to Internet advertising recommendation system 、UCG Image and video filtering 、 Medical image recognition 、 Street view, street sign identification, etc . intelligence OCR Identification belongs to multi class classification problem , Scene is complicated 、 Challenging ; Especially Chinese recognition , The character set has reached 20000 class , English numbers and letters only 62 class .

Deep extension technology intelligent one-stop OCR The service platform can provide multiple scenarios 、 Multilingual 、 High precision text detection and recognition services , It is widely applicable to ID cards 、 Account book 、 Bank card 、 Business card 、 Fiscal and tax bills 、 file 、 form 、 License plate and other specific recognition objects , Comprehensively assist enterprises in intelligent office , Reduce cost and increase efficiency .

Shenyanke is connected to intelligent OCR One stop service platform functions

Deep extension technology intelligent one-stop OCR Service platform industry solutions

insurance

utilize OCR Technology fully enabled premium withholding 、 insure 、 Property insurance, etc , To provide personalized services and technical support for all segments of the insurance industry .

• The mobile terminal of the mobile phone is insured

Shoot your ID card through your mobile phone 、 A driver's license 、 Driving license and other certificate information , And name 、 ID number 、 license plate 、 Plate type 、 Date of initial registration , Nature of use 、 Vehicle type and other information are automatically entered by category . Insurance is more convenient , Improve customer conversion .

• Customer premium withholding

The salesman can automatically identify the customer's bank card through mobile phone or tablet photography to verify the bank card information of the applicant's account , Reduce the error rate of manual input , Convenient operation 、 Save time , Improve business efficiency .

• Vehicle insurance site investigation

When the salesman collects information on site , Use mobile devices （ mobile phone 、 Tablet, etc ） shooting , By deploying OCR The technology automatically identifies and enters the identity card of the party concerned 、 A driver's license 、 Driving license and other information . Save business time , Improve customer satisfaction .

Bank

Open an account for a bank 、 Internal processes 、 Loans and other businesses create a new service ecosystem , Reduce service costs , Enhance the user experience , Create new revenue streams .

• Open an online account

Provide ocr Online identification card 、 Bank card and other opening certificate services , Enter information anytime, anywhere , More convenient 、 Effectively meet the needs of users .

• Before the loan, the enterprise made a back tone

Enabling enterprises to invest 、 Enterprises take loans and other main businesses of banks , utilize ocr Technology identification 、 Enter the enterprise three certificate information for verification .

• Internal process optimization

Realize enterprise annual report through mobile phone or tablet 、 All kinds of forms 、 Electronic documents and work order texts , Automatic entry .

negotiable securities

Assist in opening securities accounts , Account authentication , Mobile payment business , Use accurate OCR Recognition technology , Provide more intelligence for financial institutions 、 Precise and professional services , Change the business model with intelligence .

• Online account opening certification

Mobile terminal account opening authentication 、Web Account opening Certification . Support ID identification 、 Face recognition 、 Bank card identification , Precise OCR Identification technology and biometrics , Enhance the financial services experience .

• Real name authentication

Investment and wealth management customers take ID cards , The user's name can be automatically recognized 、 Id card number 、 Validity of ID card 、 Location and other information . When the bank card is bound at the same time , It can also be identified by photographing a bank card , Carry out automatic identification of card number , Easily complete the binding , The operation is more gradual .

• Mobile payment

combination OCR Certificate recognition technology and face recognition technology , Provide guarantee for the security of mobile payment identity authentication , Avoid mobile phone theft and reduce process time to improve conversion rate .

Supply chain

Accurately grasp the layout of the industrial chain , Supporting risk management 、 Marketing to expand customers or attract investment , Provide strong data support for industrial research . utilize OCR technology , Identify and input various transportation process documents and documents , Optimize process and improve efficiency .

Mobile devices

From mobile payment 、 Identity Authentication 、 Vehicle Certification , To 37 Enterprise credit data query of dimensions , Bring new experience to use .

• Mobile terminal document scanning

utilize ocr technology , The mobile end can （ Cell phones or tablets ） The device becomes a mobile scanner , Scan paper documents 、PPT、 Various bills , Generate PDF or JPG Format file .

• Identification of bank card number and other information

You can enter the bank card number by taking photos with your mobile phone , Mobile payment on mobile phones , Bank card management . And deep extension OCR Recognition technology , Not only does it support identification card numbers , You can also enter the cardholder's name at the same time 、 Card expiration date .

Smart government 、 Community

Using big data , as well as OCR technology , Promote the mobility of government and community 、 Promotion of information technology , Optimize information collection management 、 Visitor registration 、 Certificate handling and other processes , Form efficient 、 agile 、 Convenient and intelligent government affairs & Smart community .

• Government mobile office

Enterprises OA System ,CRM System and other office systems , utilize ocr Identify and quickly enter personnel information 、 ID information 、 Business card information management .

• Information service platform optimization

utilize ocr technology , Just take photos with your mobile phone , It can automatically identify and input , Complete resident identity authentication , Bank card binding , Certificate application 、 Many convenient services such as certificate replacement .

Compared with tradition OCR, Intelligent learning based on deep learning OCR Technology has higher recognition accuracy 、 Faster 、 No format dependency 、 Support privatization and rapid deployment , Deep learning algorithm and model building are also OCR The key to application . With intelligence OCR Technology is evolving , Shenyan technology will continue to refine its business needs , Combine with the existing information system , Build intelligence OCR Innovative service model , Solve the pain points in actual production .

｜ About Shenyan Technology ｜

Shenyan technology was founded in 2018 year 1 month , Zhongguancun High tech enterprise , It is an enterprise with the world's leading artificial intelligence technology AI Service experts . In computer vision 、 Based on the core technology of natural language processing and data mining , The company launched four platform products —— Deep extension intelligent data annotation platform 、 Deep extension AI Development platform 、 Deep extension automatic machine learning platform 、 Deep extension AI Open platform , Provide data processing for enterprises 、 Model building and training 、 Privacy computing 、 One stop shop for Industry algorithms and solutions AI Platform services .