CaseStudy

From Document Manager Magazine Vol 20 No 03 - May/June

In an application that won the DM Award 2011 for best Mobile Technology Project, ABBYY mobile OCR is powering a unique number plate recognition system

With the ever-growing need to efficiently control car traffic in modern cities, it is clear that automation is the only way to provide a high quality and mobile solution. For example, to monitor cars parked on city streets, police officers often write down car numbers manually or enter them into mobile phones with a device's keyboard, both of which are time-consuming and error-prone methods. At the same time, advanced software technologies today allow smartphone users to capture information with a built-in camera and instantly transform it into textual and numeric data for further processing and use.

HOW TO READ CAR NUMBER PLATES
When ABBYY Europe received a customer request for a solution that would support automated reading of data from car licence plates with a Windows Mobile smartphone, they engaged their long-term partner ATAPY Software to help develop the end solution.

ATAPY had already had a unique experience in this particular area of technology development and, by the time the project started, had already created its own Automatic car Number Plate Recognition (ANPR) software development kit (SDK).

The ATAPY ANPR SDK is powered by ABBYY FineReader Engine, one of the world's most precise Optical Character Recognition (OCR) toolkits, and contains a development library designed for reading car licence plates from still images and video streams. The SDK is a flexible solution capable of working with practically any type of camera and any Windows-based application. The product features around 90% recognition quality and is easily extendable to support new country-specific number plate types.

The ATAPY ANPR SDK was designed for operating at vehicle control points where cars slow down or stop, and is suitable for usage in a variety of applications, such as:

o Access control/gate check systems
o Parking lot management systems
o Toll road monitoring and video surveillance systems

The main challenge of the project was to port the ATAPY ANPR SDK to Windows Mobile platform, in order to enable recognition on mobile devices working in offline mode.

MOBILE OCR ON THE LOOKOUT
To address the customer's requirements, ATAPY developed a specialised mobile version of its number plate recognition toolkit by incorporating the ABBYY Mobile OCR Engine SDK. Porting a PC-based solution to a mobile environment required extra work to match the new system requirements

The number plate recognition process implemented in ATAPY ANPR SDK employed a set of sophisticated image analysis algorithms, including ATAPY's own findings, designed and optimised to address the specifics of number plate images. Specialised algorithms were used to provide smart image pre-processing and binarisation, detect the number plate rectangle on the image, and segment the rectangle into individual symbols, which were then passed to ABBYY OCR Engine for recognition. All of this was ported to the mobile platform and optimised to provide quick performance on less powerful CPUs.

Applying OCR technologies to "real world" images added further challenges to the project, for example adjusting the skew of license plates. Unlike capture by static cameras, capture by mobile devices results in images being taken from a variety of angles. ATAPY developed a specialised algorithm of automatic skew correction, capable of compensating angles of up to 30 degrees.

Additionally, the ATAPY ANPR SDK was enhanced to support a wider variety of country-specific licence plates. To implement support of a new number plate type, the engineers analysed several hundred car plate images and trained the ANPR engine accordingly.

DATA PLUS MOBILITY EQUALS LESS TIME AND EFFORT
The result of OCR processing with the ATAPY ANPR SDK is a text string containing the recognised number plus the original image of the number plate. Furthermore, it provides additional information including the coordinates of characters and plate borders, recognition confidence level (given as a percent), and other data.

The OCR results may be used for matching against a database to find out if the vehicle is stolen, or has access rights to restricted areas, etc., depending on the goals of the external application. The mobile version of ATAPY ANPR SDK is now operating as a module inside the customer's parking enforcement application, forming an essential part of a new, optimized user process.

BENEFITS OF THE MOBILE SOLUTION
ATAPY's mobile ANPR solution with ABBYY's OCR can bring many benefits to end customers. They enjoy considerable time and cost savings as a result of automating the car number input process. They achieve much higher accuracy of inputted data by using OCR instead of manual entry. They also have the advantage of 100% mobility: officers are able to work offline with a simple user interface, using the mobile device as an all-in-one workstation which provides image capture, recognition, and a number of other important functions.

"The mobility wave is opening new ways for capturing data and processing of documents," commented Bodo Wagener, VP of Sales for ABBYY Europe.

"Organisations from sectors like insurance, healthcare, transportation, banking and government, are evaluating how to integrate mobile devices into their existing processes to provide better customer service in a shorter timeframe and reduce overall operating costs. By capturing information at the point where it originates and delivering it to an enterprise application for processing enterprises can accelerate the processes, speed up transactions and reduce operating costs."

More info:www.abbyy.com

CaseStudy