Unreadable
Obscured, damaged, missing
Bent plates, mud, trailer hitches, bike racks, missing front plates and frames that cover characters. Each needs a judgment call and a defensible reason code, not a guess.
Manual image review
The transactions your ALPR gave up on.
Automation reads the easy plates. What is left is obscured, angled, glared, tailgated, temporary or simply out of state — and it is where your leakage lives. Accutech reviews that remainder, 15 to 20 million images a month, at 99.97%+ accuracy.
99.97%+Review accuracy
4×Independent reads per plate
15–20MImages reviewed monthly
24/7/365Queue coverage
Why this work still exists
Better OCR does not shrink the problem. It concentrates it.
Every point of automated read rate you gain removes an easy plate from the queue and leaves a harder one behind. The residue that reaches manual review today is worse than the average transaction was ten years ago, and the cost of getting it wrong has gone up as more of your revenue moves to pay-by-plate.
Jurisdiction
Specialty and lookalike plates
An ever-growing list of specialty designs, university and military series, and out-of-state plates that read alike at speed. Misidentify the state and you bill the wrong registered owner.
Capture
Glare, angle and tailgating
Retroreflective washout at night, straddle and tailgate captures with two plates in frame, motion blur. Often a camera or lighting problem we can trend and hand back to you.
The method
Six stages. Two independent reads before any plate is released.
Images below your configured OCR confidence threshold arrive in our queue. They leave with a disposition, a reason code and a complete action history attached to the transaction.
01
Intake and ROI crop
Transactions route to us on your threshold, not ours. We extract the region of interest from the full capture and select the best available frame where multiple exist.
02
Image enhancement
Contrast, brightness, deskew and glare handling to recover characters from a marginal capture. Enhancement is applied to the working copy; your original image is never altered.
03
Jurisdiction review, double-blind
Two reviewers independently identify the issuing state or province. Neither can see the other’s entry. Jurisdiction is reviewed before characters because getting the state wrong invalidates everything downstream.
04
Plate review, double-blind
Two further reviewers independently key the characters, validated against your configured character set, plate types, prefixes, suffixes and permitted special characters.
05
Supervisor adjudication
Any disagreement between blind reads, and any plate proposed as unreadable, escalates to a supervisor for final disposition. There is no path by which a single reviewer’s entry becomes a billable transaction.
06
Disposition, coding and return
Every transaction returns with a disposition and, where not billable, a reason code from a set we configure with you. Exempt, trusted and sensitive list matches are applied automatically and never enter a general queue.
Reason coding
A coded-off transaction should tell you something.
An unreadable plate is lost revenue either way. A reason code turns it into information — a lighting fault at one gantry, a camera drifting out of alignment, a growing share of temp tags in a corridor. We report these as trends, not just totals.
CodeApplied whenReported
NO_PLATENo plate present, or fully obscured in every available frameDaily
ILLEGIBLEPlate present, characters unrecoverable after enhancementDaily
GLARE_NIGHTRetroreflective washout — flagged back to you as a capture issueDaily + trend
ANGLE_BLURMotion blur or capture angle beyond recoveryDaily + trend
MULTI_VEHTailgate or straddle capture, more than one plate in frameDaily + trend
TEMP_TAGTemporary, paper or dealer tag, routed per your jurisdiction rulesDaily
EXEMPTMatched to your exempt or trusted vehicle list, auto-dispositionedDaily
ESCALATEDBlind reads disagreed — supervisor disposition attached to the recordReal time
Illustrative. The code set is configured to your business rules and your back office’s expected values.
Quality and audit
99.97%+ is a measurement, not a marketing number.
Accuracy that cannot be audited is an assertion. Ours is produced by the same system that runs the queues, sampled continuously, and reported to you on whatever cadence your contract sets.
You can assign your own audits against our reviewers at any time. If your quality team wants to sample our work independently, we would rather that than not.
- 01Continuous sampled audits, assignable to any reviewer or queue
- 02Per-reviewer accuracy, throughput and escalation-rate scorecards
- 03Error categorisation feeding targeted retraining, not blanket refresher courses
- 04Transactions flagged for training without leaving the production record
- 05Complete action history per transaction, exportable for your own audit
- 06Reason-code trending returned to your operations team
Capacity and service levels
Backlogs are a staffing problem. We are the staffing.
400+ trained reviewers across our US and India operations. Queue depth moves, we move reviewers onto it — weather events, incidents and camera faults absorbed inside the existing agreement.
Oldest-pending is the number that matters, not the daily average. We work to the turnaround and aging targets your contract sets, and report against them rather than around them.
If you are already behind — a system migration, a vendor transition, a camera programme — we can stand up dedicated capacity to clear the backlog while your steady-state queue keeps running.
How we contract
Three ways to bring us in.
- Prime — direct to your agency. Accutech contracts directly for image review and adjacent back-office scope, and carries the service level itself.
- Subcontract — through your integrator. Certified MBE and SBE participation through NCTRCA, the City of Houston and NMSDC, counting toward DBE, HUB and supplier-diversity goals.
- Staffing — your platform, our people. Reviewers, auditors and supervisors trained on your system and embedded in your operation, under your management.
Where your program requires review performed onshore, we will stand that capacity up in Texas and write it into the contract. Data residency, access scope and retention are documented before award — see Security & Privacy.
Send us ten thousand rejects.
The images your system coded off. We will review them, code them and score them at our cost, then show you our accuracy report next to your own numbers.