RoadPixel have a UK developed ANPR Recognition Engine but will also use other highly respected recognition engines where the situation requires it.
The recognition process comprises several stages:
- An advanced plate finder which locates plates on fast moving traffic.
- A character segmenter which identifies individual characters on the plate.
- And finally the OCR algorithms which use various techniques including neural net algorithms to interpret the plate.
The RoadPixel ANPR Engine captures plate images at a very high rate, dependent upon the complexity of the image. This throughput means that the PC can process several images of each plate as a vehicle approaches to aggregate an accurate result.
The ANPR engine has a robust level of filtering which increases the raw recognition rate well above many other systems. In addition, the user can specify a recognition zone which only searches for plates within a specific area of the screen. This filters out traffic travelling in the other direction – and removes artefacts such as may be found on roadside railings etc.
A minimum size of plate is then applied (user definable) which ignores plates below the optimum size for good recognition. This ensures that only readable plates are presented for processing. The plate result is then passed to an application program, for example RPX-LIVE, our core ANPR database program. RRX-LIVE then takes the appropriate action – such as displaying a driver’s details, controlling automatic barriers or an external LED sign – all details are saved to an event file on disk for later searching and retrieval.