Any Plate from Any State: LPR Plate Detection

Part 2: Plate Detection

It’s easy for us to identify a license plate – it’s obviously the rectangular thing on the bumper of a vehicle that contains numbers and letters. But to a camera, there might be numerous other rectangular things in the camera’s field of view that contain numbers and letters. Bumper stickers and street signs, for example. How does a camera distinguish a license plate from the surrounding environment? To do this, an LPR camera relies on certain tools. Let’s take a look at a few of the tools featured in TagMaster North America’s CS camera line.

Plate Localization/Vehicle Motion Detection (VMD)

Of course, an LPR camera’s first task is finding the plate. Many cameras are unable to identify license plates without the use of presence detection to trigger the read. This requires the installation of a ground loop. However, TagMaster North America’s LPR software and processing onboard our intelligent LPR cameras boast a feature called plate localization through vehicle motion detection, or VMD. With this feature, the LPR camera will begin reading the plate when it detects the motion of a vehicle, without requiring additional infrastructure costs in cutting additional ground loops.

With plate localization, when a vehicle drives into the operator-defined field-of-view, the camera detects the plate based on three things: the plate’s size, its shape, and its reflectivity. In addition to eliminating the need to cut additional ground loops, this feature means the camera is always scanning the environment for valid plate reads according to the directionality and distance configured for the specific site. This increases the accuracy with which license plates are recorded, and it also means license plates that do not already exist in the camera’s database (meaning the camera is not pre-programmed to recognize that particular string of characters) will still be recorded.

Character Segmentation

Once an LPR camera detects a license plate, the next step is to detect the characters the plate contains. Character segmentation is a feature of TagMaster NA’s LPR software and processing onboard our intelligent LPR cameras that allows them to isolate each character on the license plate. Isolating each character means the camera can identify and process each character individually, based on its state-prioritized syntax and the extensive OCR library (their inventory of recognized characters – read more about OCR here). This feature optimizes the accuracy with which your TagMaster NA LPR solution records each license plate ID.

DeSkewing

Vehicles drive into lots and garages at different speeds, angles and heights. This means that when a camera identifies a license plate, that plate might be at a slightly different angle than the one before it, which causes the license plate characters to look skewed. This can present an additional challenge for LPR cameras when they’re trying to match license plate characters to the characters stored in their OCR library. As you can imagine, a skewed M, for example, will look different than an M that is viewed head-on

TagMaster NA’s intelligent LPR cameras use onboard processing (in conjunction with our JetManagementSuite (JMS) software or a third party access control panel) to accommodate these angle and height discrepancies with the deskewing feature. With deskewing, the LPR camera is able to apply its algorithm to digitally “right” the license plate angle before segmenting and processing the characters. And all of these digital adjustments happen quickly enough to capture moving license plates with optimal accuracy.

Want to learn more about our LPR solutions? Visit our LPR page, or contact a TagMaster North America representative for more information.