License Plate Recognition, a Twenty-First Century Fact of Life

Article contributed by Virginia Fair

Terrorists were the intended targets for the first license plate readers deployed by New York City.  It was 2006 and the NYPD was involved in what was known as the Lower Manhattan Security Initiative, a counter-terrorism plan that involved setting up movable, random roadblocks in the Financial District. Thousands of cameras provided ancillary surveillance in the area south of Canal Street but the program revolved around special ones equipped with license plate reading technology.

Thank goodness the NYPD has been successful so far in quelling terrorist plots. They have expanded their use of license readers to attack everyday crime wherever it may be happening. According to an article in the New York Times, as of April 2011, New York was using 238 license plate readers. Of these 130 are mobile, mounted on the backs of police cars that might be patrolling any street in the city’s five boroughs. The other 108 are fixed posts at city bridges and tunnels, as well as above other thoroughfares. License plate reading cameras differ from other surveillance IP cameras that monitor broad areas in that they are designed to focus on a small area, and are aimed low to the ground.

 Working with the readers from 2006 to 2011, police tracked down 3,659 stolen vehicles, and issued traffic tickets for 34,969 un-registered ones. In the period from 2010 to 2011 alone, they identified and recovered 248 vehicles bearing stolen license plates.

Divisions dealing with felonies have used the technology to their advantage as well. In 2011 a bank robber was apprehended after high-jacking a livery cab in New Jersey and driving it through the Lincoln tunnel to New York. Somewhere along the route, the license plate was detected and the car traced to a specific block in Queens. FBI agents, alerted by the NYPD, surveyed the block and the next morning apprehended the suspect who had a loaded pistol in his possession. 

In another case of violent crime, a murder suspect was arrested after several cameras spotted his plates in various locations. The police had but to connect the dots to find him sequestered in a closet in a relative’s home. 

How does this work in a city measuring 304.8 square miles (or 468.9 square miles if one counts the 165.6 square miles of water)? The data captured on the cameras are continuously checked against specific databases containing information on stolen vehicles, stolen license plates, and unregistered vehicles. In addition, the cameras’ files are downloaded twice daily to central computers where personnel update the databases each time. Investigators are then able to retrieve new information such as the license plate of a new suspect or the stolen license plate of one they’ve lost track of.

Kintronics just introduced the TruViewLPR, license plate recognition system.  It uses special software developed about 15 years ago with the sole purpose of automatically reading license plates.   The TruViewLPR system can be broken down into four functions:
·         Image collection
·         Image analysis
·         Image and data storage
·         Data transmission

Image Collection
License plate capture cameras with CCD image sensor works with a pulsed infra-red light source to monitor a target area of passing vehicles. The illumination device contains up to 190 LEDs in the near infrared range and is capable of providing a high contrast black and white image similar to the image below.

Notice how the use of infra-red light suppresses most of the surrounding detail and allows the reflective license plate properties to make it dominant in the field.  In addition the TruViewLPR license plate capture camera lets the user alter the contrast by changing each video field up to sixty times per second, on a cycle of three different levels of brightness  - low, medium, and high. Taken together, these allow for optimal plate image processing no matter what the time of day or the condition the license plate in question.

Image Analysis
The captured images are processed by a set of algorithms that extract only the license plate portion of the frame and send it to two different Optical Character Recognition engines for processing.  It takes 200 milliseconds or less for the LPR Software to analyze and come up with an ALPR result. It then reports one of two reads: The read that provides the highest confidence score level of all the captured images for that particular license plate or the read that meets a pre-determined minimum level of confidence

Data Storage
The image with the best results is now saved and linked with the results data. The data might consist of the plate number, the date and time, the lane number.  All this information can be past to compatible Video Management Software such as Ocularis from OnSSi.

Overview Camera
In addition another camera may be used to furnish a scene overview showing a full view of the vehicle which will be linked to the plate data and image, all to be stored to be made available for subsequent  queries. You can also add many IP cameras for multiple overviews when using VMS software such as Ocularis.
 

Data and Image Management and Display
Stored data can be forwarded to a central server over a standard TCP/IP connection or using a wireless connection.

The LPR information can be displayed using Ocularis VMS software or using the Central Management console which will allow an operator to bring up ALPR events based on license plate number, date, time, lane, or other desired characteristics.

Applications
There are a number of applications where automated License plate recognition can be used.  Image collection can take place in a triggered or non-triggered environment.

·         A non-triggered installation needs no detection device. In this mode, software, known as Virtual Vehicle Detector, analyzes each image at a rate of sixty images per second for the presence of a license plate. This image, and additional images containing the vehicle’s license plate data is captured and processed to extract the license plate characteristics

·         A triggered mode requires a detection device and can be used in a number of applications. The trigger could be an in-ground loop or an optical trigger and is called for when several systems are to be tied together to a single event. Such parallel systems might be a vehicle classification system, a transponder system, a parking lot ticket dispenser, a weigh-in motion system , and so on.

The LPR Software device can act as a lane controller, hosting a database that will permit or deny vehicle access into or out of a parking facility, gated community, or high-security compound. This can be done with the optional Universal Interface Controller (UIC) to provide contact closure outputs to open or close a gate or arm in response to queries of the database.

And so in this day and age, a license plate serves as more than just a way to determine if that’s your buddy in the silver Honda up ahead. 

For more information and help with your system design, please contact us at 914-944-3425 or 1-800-431-1658 (in the USA), or you can use our contact form.