LPR matcher technique: OCR equivalence

The LPR matcher uses the Optical Character Recognition (OCR) equivalence technique to improve plate read accuracy rate.

Depending on the font design, some plate characters can look very similar to other characters. These are called "OCR equivalent characters".

You can configure how the LPR matcher handles OCR equivalence by modifying the MatcherSettings.xml file. For more information, see MatcherSettings.xml file.

NOTE: You can also configure the following LPR matcher techniques in the MatcherSettings.xml file:

Number of character differences (see LPR matcher technique: Number of character differences)
Common and contiguous characters (see LPR matcher technique: Common and contiguous characters)

The default Latin-based OCR equivalent characters are the following:

The number “0” and the letters “O”, “D”, and “Q”.
The number “1” and the letter “I”.
The number “2” and the letter “Z”.
The number “5” and the letter “S”
The number “8” and the letter “B”.
The number “6” and the letter “G”.

BEST PRACTICE: You should not allow more than two OCR equivalent characters because it results in too many false-positive matches.

Example

The following example uses a hotlist with the LPR matcher configured to allow for one OCR equivalent character:

The LPR matcher finds the exact match ABC123 in the hotlist and raises a hit. It also looks for any plates that are one OCR equivalent character off, and finds A8C123, ABC1Z3, and ABCI23 in the hotlist, so it raises hits on them as well.

If the LPR matcher found the plate A8CIZ3 (three OCR equivalent characters off), it would not raise a hit because the system is configured to accept a maximum of one OCR difference.