ABBYY Real-Time Recognition

ABBYY Real-Time Recognition - mobile software for text recognition from any surface. It is expected creating applications, the mobile devices considering restrictions of RAM.

2017

Partnership with VisionLabs

In September ABBYY provided two products for remote identification of users. With their help it is possible to recognize "on the fly" passport details, to confirm the client's personality on a selfie and to verify authenticity of its documents. Together solutions provide the safest remote check of the identity of the user. Thanks to it it is possible to register far off clients, to open accounts, to make out transactions and to confirm transactions, without risking to make such operations with the counterfeit or stolen document.

The first solution is created by ABBYY in partnership with VisionLabs company, software developer for face recognition. For remote identification it is enough to client to point the device camera at the passport or other identity certificate and to look in the camera. The ABBYY Real-Time Recognition technology takes the photo and will recognize the text from the document. The VisionLabs Photo Maker technology which is a part of the solution allows to select for several seconds the best picture of the person from a video flow. For increased security also Liveness the detector is built in a product. During check a system suggests the user to blink and analyzes a video flow, preventing an input in mobile application of bank or confirmation of transaction by the counterfeit person – on a photo or the recorded video with the client's person. The solution compares the user's person to his photo in the document and defines percent of similarity of two images. At positive result a system sends the notification on successful passing of check.

Also ABBYY provided a prototype of the solution which verifies authenticity of the passport or other identity certificates. The prototype is developed in partnership with the specialist company in the field of technologies for a digital security of business. For check passing the client needs to point the smartphone camera at the page of the document. The technology for several seconds takes images and the text, and then depending on document type sends it for check by a number of text and graphic parameters. For example, a system can determine counterfeit of the passport by the color ornament printed in the photo and patterns on pages of the document. Automatic loading of results of verification and data on the client in information systems is possible.

Products can be used separately and in a uniform complex. Together solutions will help to provide quickly and safely to clients services which require preliminary verification of documents.

See Also:

• ABBYY PassportReader
• National Biometric Platform (NBP)
• Technologies of biometric identification
• Biometrics and border control
• Sonda RTI
• IDensic
• VisionLabs
• Vocord FaceControl
• Biometric identification
• Oz PhotoExpert
• Gemalto Coesys Border and Visa Management

SDK and DDK sets

On March 29, 2017 the ABBYY company announced release of tool kit for developers which will allow to create applications on the basis of ABBYY Real-Time Recognition – sensing technologies of the text taken by the camera practically from any surface.

The technology can convert into the machine-readable text contents of documents, labels, meter readings, car numbers, etc. Text recognition from screens of different devices^[1] is possible^[1].

It is the only product in Russia which "on the fly" finds and will recognize texts on any documents and surfaces of objects. Dmitry Shushkin, deputy CEO of ABBYY Russia

According to the statement of the company, the technology is used in public sector, the sphere of logistics and a telecom of the Czech Republic, Brazil, China, Sri Lanka and Cambodia. In Russia it attracted interest from banking, insurance companies and retail.

How it works

For text recognition the device camera with software of ABBYY RTR should be pointed at an object with a text. The technology will fascinate the image and will recognize characters. It is not necessary to take a picture - ABBYY RTR will work at guidance of the camera. Developers promised rather high accuracy of recognition. Image capture in a video flow excludes influence of defects of photographing on recognition accuracy.

In the course of recognition the taken character is compared to letters and digits. All recognizable combination is compared with the dictionary of the alphabetic and numeric expressions characteristic, for example, of documents. It helps to avoid widespread errors, the and type instead of "й", not to confuse letters, similar on outlines, and digits, it seems "8" and "In". For simplification of recognition in a video flow excess elements are eliminated. For increase in accuracy the received result undergoes testing of a format. For example, date should correspond to a format 0/0/0000.

The procedure of recognition does not require connection to the Internet as it is implemented on the device, but not external servers. According to ABBYY it reduces risk of date leak.

The technology will recognize texts in 63 languages, from 9 languages the translation into English and back is possible. Among these languages:

• Indonesian,
• Spanish,
• Chinese,
• German,
• Polish,
• Portuguese (Brazilian),
• Russian,
• French
• Japanese.

ABBYY created the ABBYY Real-Time Recognition SDK tool kit, it will allow third-party developers to create on the basis of this technology applications for iOS and Android. ABBYY provides to developers the license to use SDK in the client's product.

According to developers, these tools will be demanded in the field of service. In addition to banks the application can be demanded in insurance companies, retail chain stores, hotels and hotels, in the field of telecommunications.

Representation of Real-Time Recognition Mobile Capture SDK from ABBYY, (2016)

System Requirements

Supported operating systems

• iOS 7.x above
• Android 4.4 above for ARM v7 of processors

Processor

• Multi-core processor
• Advanced SIMD (NEON)

Memory

Consumption of memory at text recognition:

• for the European languages — 40 MB of RAM
• for hieroglyphic languages — 70 MB of RAM

Camera

• Auto focus
• HD предросмотр: the recommended size 720*1080, can change depending on the scenario and processing speed

Notes