# Vision measuring systems

Camera-based coordinate measurement, lighting and edge strategy, autofocus, performance specifications, and application qualification.

Reviewed: 2026-11-01

Tags: inspection, dimensional-metrology, optical-measurement, vision

## Measurement principle and scope

Vision systems combine imaging optics, a camera, coordinates and software. Edge detection and alignment settings affect the result. [Mitutoyo: Quick Guide, vision-measuring section](https://www.mitutoyo.com/webfoo/wp-content/uploads/E11003-Quick-Guide.pdf)

**Illustrative sensitivity check:** suppose a locally qualified width routine gives 10.004 mm under one lighting setting and 10.012 mm under another. The 0.008 mm difference calls for investigating boundary selection and settings; averaging the two is not a demonstrated correction. These invented readings illustrate a program-validation question, not typical system performance.

Use the system for a defined measurand: a width between selected edges, a diameter fitted to sampled boundary points, a position relative to datums, or a height difference measured with a qualified sensor. Record the viewing direction and feature definition. Visible texture, a burr, a coating boundary, and a machined edge may all generate contrast without representing the same drawing feature.

## System configuration matters

| Configuration item | Question for selection and qualification |
| --- | --- |
| Optics and magnification | Can the system resolve and identify the required boundary over the selected field? |
| Stage and coordinate system | Does the specified performance cover the working volume and measurement direction? |
| Illumination | Which light source and settings reveal the intended feature consistently? |
| Focus or height sensor | Is the method appropriate to the material, surface contrast, slope, and height range? |
| Software | How are points selected, filtered, fitted, and related to datums? |
| Fixture | Is the component supported in its specified condition without obstruction or distortion? |

Manual systems require operator movement and focusing; CNC systems execute programmed motion, focus, and measurement routines. Mitutoyo's product guidance describes both, along with systems combining vision and contact measurement. The applicable configuration must be identified rather than treating every camera-equipped machine as the same instrument. [Mitutoyo: vision-system guidance](https://measure.mitutoyo.com/catalog-vision-measuring-systems/)

## Illumination and edge strategy

Transmitted lighting reveals a silhouette; coaxial lighting illuminates along the optical axis; ring lighting provides selectable illumination angles. Mitutoyo identifies these as different tools for generating usable feature contrast. [Mitutoyo: lighting options](https://measure.mitutoyo.com/catalog-vision-measuring-systems/)

Specify lighting as part of the measurement program. A stable algorithm can repeatedly select the wrong boundary if the region of interest includes a reflection or adjacent edge. Challenge the program with representative surface finishes, colors, burr conditions, and allowable part variation. Verify which side of an edge transition is detected and how the feature is fitted.

Subpixel estimation is a processing technique, not an independent accuracy certificate. An algorithm may estimate an edge between pixel centers, but lens behavior, calibration, illumination, and the physical edge still determine whether that estimate corresponds to the required coordinate. Avoid selecting acceptance limits solely from pixel count or image enlargement.

## Autofocus and Z measurement

OGP's *Video Autofocus* technical report, **part 791083-0324, copyright 2024**, explains autofocus as evaluating image contrast while stepping through focus. It identifies optical depth of field, light source and brightness, focus-tool area, and focus strategy as important influences. A focus region should not overlap surfaces outside the plane of interest. Its Z-verification discussion uses a suitable step artifact covering multiple heights, rather than inferring height accuracy from a single sharp image. [OGP: Video Autofocus report](https://www.ogpnet.com/wp-content/uploads/OGP_Tech_Report_Video_Autofocus_OGP-NA_791083_0324_LR_SECURED-1.pdf)

Define whether Z comes from video autofocus, a contact probe, or another optical sensor. Confirm repeatability on the actual surface and range. A reflective or low-contrast surface may require a different acquisition strategy; a large Z travel specification alone does not establish usable optical height measurement on every material.

ZEISS's **O-INSPECT** product page distinguishes its **Discovery.V12** camera, **VAST XXT** contact sensor, and optional **DotScan** topography sensor. It also identifies ground vibration as a possible source of erroneous measurements. A multisensor machine offers alternative methods, each requiring its own applicable performance and setup. [ZEISS: O-INSPECT sensors and environment](https://www.zeiss.com/metrology/en/systems/cmms/vmms/o-inspect.html)

## Bounded manufacturer specification example

Mitutoyo's **Quick Vision Pro bulletin 2316, printed January 2022**, lists **QV APEX 302 Pro** with **300 × 200 × 200 mm measuring range** and **0.1 μm scale resolution**. Its vision length-error specifications distinguish **EUX/EUY,MPE = (1.5 + 3L/1000) μm**, **EUXY,MPE = (2.0 + 4L/1000) μm**, and **EUZ,MPE = (1.5 + 4L/1000) μm**, with L the measurement length in millimetres. [Mitutoyo: bulletin, page 18](https://www.mitutoyo.com/webfoo/wp-content/uploads/Quick_Vision-Pro.pdf)

At L = 100 mm, the listed XY expression evaluates to 2.4 μm. That is a manufacturer permissible-error expression under the applicable conditions, not the uncertainty of an arbitrary part measurement. The brochure discusses temperature compensation and its accuracy-guarantee conditions separately; APEX compensation is manual. Confirm the supplied machine, optics, selected verification package, current documentation, and environmental conditions before applying a specification. [Mitutoyo: bulletin, page 16](https://www.mitutoyo.com/webfoo/wp-content/uploads/Quick_Vision-Pro.pdf)

## Calibration and application qualification

ISO **10360-7:2011**, published June 2011, specifies acceptance and reverification tests for manufacturer-stated linear-dimension performance. Its public scope applies to **Cartesian CMMs with imaging probing systems operating in discrete-point mode**. It does not automatically cover every optical measurement technology or contact-probe function. Match any compliance statement to the actual sensor, mode, and test scope. [ISO: ISO 10360-7 scope](https://www.iso.org/standard/43904.html)

Machine verification and application qualification serve different purposes. Verify the relevant coordinate and sensor performance with appropriate calibrated artifacts, then evaluate the actual inspection routine. Include different stage locations, feature orientations, magnifications, refocusing, and reloading where these occur in use. Establish datum construction and fitting rules before using the program for acceptance. Repeating the same saved image tests software repeatability but excludes new acquisition and fixturing effects.

NIST's **8 October 2026** publication on its upgraded **M48 vision measurement system** combines accurate CMM positioning with camera-based edge detection. The study validates proposed uncertainties using long-term check standards and reference materials and explicitly considers the vision artifacts being calibrated. Its uncertainty results belong to that system and study, not to commercial vision machines generally. [NIST: M48 validation study](https://www.nist.gov/publications/validation-upgrades-nist-m48-vision-measurement-system)

## Reporting and interpretation

Assess uncertainty from coordinate performance, optical calibration, edge localization, focus, alignment, fitting, fixture behavior, temperature, and repeated acquisition. Separate display increment, repeatability, permissible error, and task-specific uncertainty. Automation can reproduce a consistent bias as readily as a correct measurement.

A useful record includes system and sensor identity, optics, program/software revision, lighting and focus settings, fixture and datum definitions, sampled regions, filtering and fitting, reference checks, environmental conditions, results, and the applicable uncertainty and acceptance rule. Preserve enough configuration information to reproduce the result when a supplier or later inspection disagrees.

## Related reference

[Inspection and quality documents](../../sourcing/documents/index.md)