Plate-confidence settings need an understood score and an authorised review process, with capture problems kept visible during testing.
Thresholds should be controlled by authorised people who understand what the product's score measures and how changes affect review. The supplier needs to explain the available settings; the organisation needs an agreed way to test, approve and record changes.
A higher threshold does not establish that every accepted match is correct. Scores from different engines may describe different things, so copying a number between products is not a meaningful substitute for examining the records.
Recognition software developed from producing a text candidate to supplying scores and review filters alongside it. A score reflects an engine's assessment within its own model, rather than a universal probability that a vehicle has been correctly identified.
Image quality, individual characters and whole-plate matching can each have a score. Before changing a threshold, establish which stage it governs and whether another filter has already removed observations earlier in the data path.
Thresholds matter when uncertain observations need a defined review route or a trial shows relevant records being excluded. Sometimes illumination or capture position is the more relevant change, especially where the image itself contains too little detail.
An assistance process or another credential may be appropriate for vehicles the camera cannot observe adequately. A missing exit also needs session reconciliation; treating every open visit as a confidence-setting problem can hide that distinction.
Use a labelled sample containing ordinary observations, ambiguous text and unreadable plates. Record what the proposed setting excludes, admits or sends for review. Keep the original images and settings so a supervisor can understand why each record followed its path.
Keep source-engine thresholds separate from downstream review rules. If a camera discards an event before sending it, lowering a CORE filter cannot recover it. The test needs to show both stages rather than only the final review queue.
Aero Ranger CORE's Detection Confidence feature describes organisation-defined thresholds, visible exceptions and authorised review. It is relevant to controlling which records reach review while keeping the score separate from the eventual decision.
The source score, available threshold scope, discarded-event handling and change permissions still need confirmation. Those details explain what a configuration can affect and what remains a capture or session-reconciliation issue.
Aero Ranger: Detection confidence
Sources were reviewed on the publication date. Check the linked material for current details.
