Part 1: Character Recognition
1’s and I’s, S’s and 5’s: How Do LPR Cameras Distinguish between License Plate Characters?
Have you ever had trouble determining if a license plate character was an I or a 1? An S or a 5? An O or a 0? If we have trouble making such distinctions, how do LPR systems do it?
While cameras may lack the critical thinking skills that help people distinguish between similar-looking characters, LPR cameras have other powerful tools at their disposal that help them make such distinctions. Let’s take a look at some of the tools TagMaster North America’s CS family of cameras use to achieve accurate plate reads.
Optical Character Recognition
The optical character recognition (OCR) library is arguably the most important factor for LPR accuracy. The OCR library is the inventory of characters your camera recognizes – every letter and number and its numerous potential variances in fonts and shapes. If your camera has a limited OCR library, it has to rely on its ability to identify characters that look similar to the characters it’s learned; this can lead to incorrect vehicle identifications if, for example, your camera sees an I and mistakes it for a 1. Therefore, the more extensive the OCR library, the more reliable the license plate recognition.
TagMaster NA’s family of CS cameras operate with an OCR library that is owned and managed by TagMaster. Our OCR library can be updated to incorporate new plate releases, syntaxes and reflectivity, and it is supported by a team of PhD software developers who take real-world feedback and apply it to how our LPR cameras “think.” This ensures our CS cameras have an extensive OCR library and can be counted on to accurately interpret the license plate characters they record.
State by State Syntax
There are a lot of variables in the appearances of North American license plates. With no federal regulation, every state’s plates can use different formats, fonts, colors and graphics. One state’s plates might follow the format ABC 123, whereas another state might follow the format 1ABC234. And one state’s B might resemble another state’s 8. With all of these variances, an LPR camera needs to be highly intelligent to accurately identify license plate characters.
Many LPR cameras are limited in the number of fonts and syntaxes they can recognize, because the amount of information and variations simply become too much for these cameras to support. Ideally, a high quality LPR solution can provide your camera with an extensive inventory of characters and syntaxes, accounting for the numerous differences that can occur between and within states. TagMaster NA’s LPR solution does just that. Our family of CS cameras has a feature called state-by-state syntax, which provides an exhaustive database of potential fonts and formats, so your camera can accurately read and record state-issued license plates from anywhere.
- Plate 1 is from the state of Oregon, with the ID 789 ABC.
- Plate 2 is from the state of California, with the ID 6ABC789.
Now imagine Plate 1’s character for the letter B looks like Plate 2’s character for the number 8. A camera without state-by-state syntax might confuse these two characters and incorrectly record both license plate numbers. But with state-by-state syntax, TagMaster NA’s cameras can use the syntax of the license plate to determine which state the plate is from, and it can use this knowledge to determine which character, B or 8, it should record. So in this example, when TagMaster NA’s camera sees a license plate that follows the format 789 ABC, it knows the plate is from Oregon, and can accurately transcribe Oregon’s characters for 8 and B.
