- Vision System Performs License Plate Verification
Anne L. Fischer
Border crossings, secure areas, parking facilities and even highway speed traps are places where license plates on cars may be captured by imaging systems. In Italy, a high-tech license plate reader is being used by an intelligence service, an application in which a plate is read, the information sent to a database and the vehicle allowed to proceed to a secure area.
Advanced Technologies srl of Milan and Sinartis srl of Cassano Magnago, both in Italy, took up the task of designing and installing the system in just six months. License plate reading equipment is in use in many places around the world, and the companies met the key requirements for this project by using an off-the-shelf system, rather than starting from scratch.
An intelligence service uses feature-based optical character recognition to read license plates on vehicles.
The first requirement was that the license plate reader achieve high-quality imaging in a real environment. According to Davide Nardelli, managing director of Advanced Technologies, the lighting situation in such an application can be complex. "The sun could be on the front or back of vehicle; it could be completely bright or completely dark," he explained.
Another was that the system should work attached to any base and from any advantage point. For example, components could be attached to a moving gate or a fixed pole and could be exposed to the elements or shaded.
The last requirement was quick capture of a plate number and transmission to a centralized database by a wireless network. The system also should simultaneously compress and save the image locally so that it could be downloaded off-line.
The system chosen for the task incorporates a radar unit, a Basler Vision Technologies A102k 1.3-megapixel Channel Link camera, a Matrox 4Sight-II industrial vision computer and lighting. The radar locates the moving vehicle, triggering the illumination and the camera. The LED-based light source is flashed, and the image is captured. The radar and lighting system are synchronized to ensure that the license plate is in the correct field of view and properly lighted for image capture.
The 4Sight-II, equipped with a Matrox Meteor-II/Camera Link frame grabber, performs image acquisition, processing and display functions. The designers elected to use Matrox Imaging Library software with a feature-based optical character recognition tool that can handle extreme contrast variations, such as from day to night, and geometric distortions. The application also uses the software's compression/decompression module, featuring an optimized jpeg compression algorithm.
Using the Matrox industrial vision computer, rather than another off-the-shelf product, ensured compatibility among the elements of the system. This reduced the amount of time that Advanced Technologies had to spend confirming component interoperability, both during development and once the system was put into use. Another advantage of the computer was its small size, enabling it to fit into spaces that wouldn't accommodate a traditional PC-based vision system.
The advantages of using integrated software included keeping down costs and ensuring that the system would not require recalibration when moved to a different light pole or parking lot gate. The optical character recognition module offers a common programming interface that is compatible with Matrox hardware, and Advanced Technologies found it ideal for its application, in which the font and the location of the string of characters are unpredictable.
A helpful feature in this application was the ability to specify a set of "grammar rules" that direct the system to look for a series of characters on the license plates and to respond to specified combinations.
Regarding the requirement of speed, the system can perform plate recognition and image compression on a 1300 × 1024-pixel-resolution image in less than 500 ms. That is the minimum requirement to ensure that moving vehicles are identified, Nardelli said. The application designers determined that the minimum CPU speed necessary to process all tasks was 1 GHz.
The intelligence service seeking the system also wanted an open programmable environment with the right flexibility for customization. By working with Matrox, this effort helped drive research and development on the company's feature-based optical character recognition software. This level of programmability is included in the Matrox Imaging Library string reader module.
Although the system meets the requirements for the application, subsequent iterations will include a lower-resolution camera. This change is mainly to cut overall costs, and the system designers are confident that image quality will not be greatly sacrificed. Other changes may include using a smart camera such as the Matrox Iris. According to Nardelli, the advantages of using a smart camera with this system are initial cost reduction, ease of integration and savings on maintenance.
MORE FROM PHOTONICS MEDIA