← All glossary terms Glossary

OCR (Optical character recognition)

Optical character recognition (OCR) is the process of converting images of letters and numbers into machine-readable text; in ANPR it is the step that reads the characters on a detected number plate.

What it means

Older OCR engines segmented the plate into single characters and matched each one against templates. Current ANPR systems use neural networks that read the whole plate as a sequence, which copes better with uneven spacing, dirt and perspective. Most systems also check the result against the plate format rules of the country, so that a read with an impossible state code is flagged.

How it applies in India

Indian plates challenge OCR in specific ways: two-line and three-line layouts on two-wheelers and commercial vehicles, hand-painted fonts, decorative fonts on older plates, stickers and dirt, and confusions such as 0 and O, 8 and B, or 1 and I. HSRP plates use a standard font, which helps. Ask a vendor how the model was trained on Indian plates, and how the read rate is defined: the whole plate correct, or individual characters correct. Those are different numbers.

Related terms

Questions about OCR on your project?

Request a demo All glossary terms