PlateScan is the only Automatic License Plate Recognition product based on a proprietary neural network recognition engine. Neural network technology offers significant benefits over template-based Optical Character Recognition (OCR) ALPR systems, since it can deliver dramatically higher performance and accuracy than other systems currently available. In mission critical situations, where accuracy and performance must be achieved and maintained, the PlateScan neural network technology is the preferred choice over other solutions.

The PlateScan Neural Network features distributed parallel processing. Therefore, it is capable of accurate high speed plate recognition across multiple lanes of high density and fast moving traffic. The software, together with an onboard processor, is currently capable of processing video feeds from four separate color and/or infrared cameras simultaneously, whether mounted on and in a patrol car or at fixed locations. It has been tested and proven to be effective at locating, scanning, and accurately recognizing license plates on vehicles at up to 110 mph .

These neural networks are not programmed in the traditional sense, but are trained by example and computer simulations on an extremely large number of repetitions of 37 character sets, covering each number, letter and a 37th for “extraneous noise” on a plate. Consequently, PlateScan’s neural network-based techniques are more able to recognize poorly defined, distorted, partially obscured and dirty plates, dramatically enhancing mission critical accuracy in the most demanding real world conditions.

When a license plate is located on a vehicle, a trigger initiates a suite of software recognition algorithms. These algorithms identify the position of the plate in the image, extract the plate image, apply geometric distortion compensation, if necessary, and segment the plate into individual characters.

Each character is then fed into the neural networks, which produce a result for every character, outputting the string of characters which makes up the license plate. The software then compares each recognized license plate to any number of local databases stored on a mobile computer, audibly alerting the patrol officer when a “hit” is registered.