Skip to content

Blog · Measurement

Dimensional gauging and tolerance checks in millimetres.

Updated July 2026 · 6 min read · Adente Vision Engineering Team

Dimensional gauging is the deterministic side of Adente Vision's hybrid approach: classical computer vision measures feature coordinates, hole positions and dimensions in millimetres per part, then checks them against tolerance. The check runs on the same unit as the four AI inspection modes, so one capture measures and judges the part.

What does dimensional gauging measure on a part?

Dimensional gauging turns an image into numbers with units: how long, how wide, how far apart, where a hole sits. Instead of asking whether a part looks right, it asks whether a feature is within tolerance, and returns a value in millimetres you can log and act on. On a real part that means linear dimensions such as a length, a width or a gap, hole and feature positions as X and Y coordinates, hole diameters, and the spacing between two features.

This is the measurement half of the hybrid design. Adente Vision is an edge-AI visual inspection unit built by ADENTE Advanced Engineering Technologies, part of the Aden Group, sold through automation system integrators, and it combines classical computer vision for measurement with AI inference for judgement. Dimensional gauging is not a separate product mode: the four inspection modes are Anomaly, Defect, Counting and Quality, and measurement is the deterministic capability that runs alongside them, using classical CV to produce feature coordinates and dimensions per part.

The unit captures the part once and can both measure it and judge it, so a single station returns a millimetre reading and an appearance verdict rather than needing a separate gauge downstream.

Why does classical computer vision give deterministic millimetres?

Measurement wants a repeatable number, not a probability, and that is exactly what classical computer vision provides. Edge detection finds the boundary of a feature to sub-pixel precision, and a calibration that maps pixels to millimetres turns that boundary into a real dimension. Give the same part and the same lighting, and the algorithm returns the same value, because it is geometry, not a learned likelihood. That determinism is why measurement sits on the classical side of the hybrid split while judgement calls, is this a defect, sit on the AI side.

The reading depends on the optics and the calibration. The unit uses up to a 12 MP global-shutter camera on a C-mount lens, so the resolution and the field of view set how fine a feature it can resolve. The accuracy in millimetres you can commit to for a given tolerance depends on that field of view, the calibration and the part presentation, so it needs an application-specific measurement rather than a single headline figure. A tolerance check is only as trustworthy as the calibration behind it, which is why measurement is set up against a known reference at commissioning.

How does dimensional gauging combine with the other checks on one unit?

The value of measurement on this unit is that it does not stand alone. Because one capture feeds both the classical CV measurement and the AI judgement, the unit can confirm a part is dimensionally in tolerance and free of cosmetic defects in the same pass. A machined housing can be measured, hole position and bore in millimetres, while the anomaly mode flags a scratch, a burr or a short-shot that no tolerance would catch. The verdict the PLC receives can combine both: out of tolerance or visibly defective both fail.

Measurement checkHow it is measuredOutput per part
Linear dimension (length, width, gap)Classical CV edge detection, calibrated pixels to mmValue in mm, pass/fail against tolerance
Hole or feature positionFeature located in the calibrated field of viewX and Y coordinates in mm, deviation from nominal
Hole diameter or boreEdge fit around the featureDiameter in mm, pass/fail against tolerance
Feature-to-feature spacingDistance between two located featuresSpacing in mm, deviation from nominal
Position or profile deviationLocated geometry compared to the nominalmm deviation, logged for traceability

Pairing measurement with anomaly is what lets one station replace a gauge and a visual check at once. The tolerance side follows the same intent as geometrical product specification standards such as ISO 1101, where position and form are defined against a nominal rather than eyeballed.

How are measurements recorded for traceability?

A measurement is most useful when it is kept, not just checked. The unit logs the per-part values and the pass/fail against tolerance to its web-based dashboard, and sends the verdict to the PLC over the fieldbus, so a dimension is both a live gate and a stored record. When a downstream problem appears, the log shows whether the feature was drifting toward its limit before it crossed, which turns a scrap investigation into a data question.

Recording measurement rather than a bare pass/fail also supports process control: a slow trend in a hole position or a bore diameter is an early warning that a tool is wearing, visible in the numbers before any part fails. For the wider design that puts measurement and judgement on one unit, see the sibling post on hybrid machine vision; to see where gauging sits alongside the other tasks the unit runs, browse the inspection applications; and for the full method behind the modes, see the pillar guide.

Frequently asked questions

Need a dimension checked in millimetres, not by eye?

Send us a sample part with the features and tolerances that matter, and we show the millimetre measurement and the pass/fail before quoting. See how Adente Vision measures and judges a part on one unit.