Fuzzy · S / 5
VendRead
8ABC12S
Paid
8ABC123
Vend. Near-miss scored against a session from two minutes ago.
LPR & matching · AI Intelligence
AI reads the plate. Then it forgives the near miss.
Detection at the lane, on Watch, and on the lot. Fuzzy matching scores 0/O and S/5 against sessions you already have. It does not invent a plate.
Three surfaces. One plate object.
Detection lives at the lane, on the phone, and on the lot. Deep links go to the product pages that already own the hardware story.
Local controller
AI detection runs at the island. Paid, permitted, or close-enough vends without a WAN hop. Fuzzy lives in that if-statement.
Gates & kiosksFuzzy match
Camera read
8ABC12S
Candidate
8ABC123
S / 5 · 0.86. Vend. Near-miss scored against a session from two minutes ago.
Vend- 0 / O
- Often the same glyph. Treated as equivalent.
- 1 / I
- Substitution against sessions and allowlists.
- 8 / B
- Common night / IR near-miss.
- S / 5
- The demo case on Gates. Still not perfect OCR.
Common questions.
Lane depth lives on Gates. Handheld detection lives on Scan to Cite.
What if LPR misreads a plate?
Fuzzy matching scores near-miss reads (0/O, 1/I, 8/B, S/5) against recent sessions and the allowlist. A hit still vends. A miss holds for QR, the kiosk, or support. It does not claim perfect reads.
Does Scan to Cite read plates fully on-device?
Detection and AUTO capture run on the phone. Character read happens on the captured still. That is on-device detection, not fully offline LPR.
Is fuzzy match a second OCR guess?
No. It scores the read you already have against sessions and the allowlist. It does not invent a new plate.
The camera can miss. The arm still goes up.
Fuzzy match at the lane. On-device lock in Watch. Identity on the stall session. Request a demo, or open the occupancy model.
- Lane
- Local vend
- Watch
- On-device lock
- Fuzzy
- Near-miss