Camera resolution for measurement accuracy in motion

A camera can provide enough pixels to show a feature without measuring it accurately. That distinction matters when the camera moves, the field of view changes, and the required error is only a few micrometres. Exposure time, shutter behavior, lens geometry and calibration can consume the error allowance before pixel count becomes the limiting factor.

For a field of view between 20 and 80 mm, a measurement target of 0.01 mm and acquisition between 10 and 16 fps, start with separate sampling, motion and measurement-error budgets. Then select hardware against a validation plan rather than a megapixel target alone.

Define what the accuracy requirement means

Before comparing area scan cameras, establish whether 0.01 mm means maximum permissible error, repeatability, measurement resolution or the smallest feature you must detect. These are different requirements.

A system might repeatedly report the same dimension while remaining consistently wrong because its calibration scale is incorrect. Conversely, a displayed increment smaller than the tolerance says nothing about measurement uncertainty.

Write down the following before selecting hardware:

• The measurand: edge position, distance between edges, diameter, gap or position in stage coordinates.
• The permitted error, including whether it is a signed limit such as ±0.01 mm.
• The full measurement range and required field of view in both axes.
• Relative motion during exposure, not just the camera's travel speed between images.
• Permitted object-height variation and camera-pose variation.
• The required acquisition rate and maximum acceptable measurement latency.

The calculations below assume a maximum permissible error of ±0.01 mm. They are screening calculations, not proof that a particular camera meets that requirement. For the broader distinction between sampling and positioning performance, see vision-guided positioning accuracy limits.

Convert field of view into a sampling budget

Object-space pixel size describes the distance represented by one image pixel at the measurement plane. Use the pixel count across the relevant image axis, not the camera's total megapixel count.

For a fixed sensor, changing the lens or working distance changes the field of view and therefore the sampling scale.

A pixel size equal to the permitted error is only a starting calculation. It does not include edge-location noise, distortion, calibration error or motion. A deliberately conservative screening option is four pixels across the permitted error magnitude; that is an engineering assumption, not a universal metrology rule.

The second column is not an accuracy guarantee. The final column shows how demanding a conservative sampling allocation becomes at the widest field. You may instead justify coarser sampling through validated subpixel measurement, reduce the field, or change the measurement method.

Use the resolution calculator to explore sampling requirements before filtering the camera finder.

Work through a candidate imaging geometry

Consider a hypothetical image with 4,000 active pixels across its width. At the narrowest field, the calculation is:

At the widest field, the same pixel count produces:

The wider field therefore requires the complete measurement to remain accurate within half a pixel. That may be possible for a stable, high-contrast edge under controlled conditions, but it cannot be assumed from pixel count or an algorithm's advertised subpixel increment.

Distance measurements also involve more than one edge. Noise, edge asymmetry and local distortion can affect the two edge locations differently. Validate the final distance calculation, not just the repeatability of one fitted edge.

As an illustrative planning budget, allocate bounded contributions as follows:

Adding these allowances leaves 2 µm below the assumed 10 µm limit. Actual contributions must be established experimentally. Do not combine unknown systematic errors by root-sum-square merely to obtain a smaller total.

The related guide to object-space pixel size from magnification explains the sampling relationship from the lens side.

Set exposure from relative motion, not frame rate

Frame rate determines how often an image is acquired. Exposure determines how far the image moves while light is being collected. A low acquisition rate does not automatically mean low motion blur.

For approximately constant transverse relative velocity:

Suppose relative speed during exposure is 100 mm/s and the sampling scale is 5 µm/pixel. If the initial blur allowance is half a pixel:

This is an exposure limit to investigate, not a prediction of dimensional error. Blur affects different edge profiles and measurement algorithms differently. Compare measurements at shorter exposures to establish whether the selected allowance is adequate.

An acquisition rate of 16 fps gives 62.5 ms between nominal frame starts; 10 fps gives 100 ms. Neither interval permits a long exposure if the camera is still moving quickly.

Short exposures require sufficient illumination and lens throughput. A synchronized light pulse can limit effective illumination time when it is correctly contained within the sensor's exposure interval and ambient light contributes negligibly.

Separate shutter, triggering and throughput requirements

For continuous relative motion, a global shutter is generally the simpler starting point. Rolling shutters expose different rows at different times, which can introduce geometric distortion even when individual rows have short exposures.

A rolling-shutter camera may still be suitable when the camera and target stop and settle before capture. Specialized illumination timing can also work in some documented sensor modes, but it must not be assumed available.

Check these timing requirements independently:

1. Exposure: short enough for the motion budget.
2. Shutter timing: appropriate for the geometry being measured.
3. Trigger acceptance: reliable at the required acquisition rate.
4. Transfer and processing: able to sustain the full acquisition sequence.
5. Position association: accurate enough when measurements are transformed into motion-system coordinates.

At 100 mm/s, a timing uncertainty of 100 µs corresponds to 0.01 mm of travel. That matters when locating a feature relative to a moving stage. A rigid image shift does not necessarily change a dimension measured entirely within the same frame.

Bandwidth must also be calculated from the delivered image format. A hypothetical 4,000 × 3,000 image at 16 fps and one byte per pixel requires:

That excludes transport overhead. Higher bit depths, packed formats and color processing can change the payload. Diagnose acquisition limitations using camera frame-rate bottleneck checks, then confirm operation with timestamped frame-rate measurements.

Check optics and calibration before adding pixels

More pixels cannot recover detail that the lens does not transmit. Nor can they correct an unstable relationship between object height and image scale. Select machine vision lenses for coverage, working distance, usable detail and measurement geometry together.

With a conventional perspective lens, object-height changes alter magnification. For small changes around a fixed setup, an approximate relationship is:

For an illustrative effective object distance of 200 mm, a measured length of 20 mm and a height change of 0.1 mm:

This first-order estimate depends on geometry and is not a substitute for a lens model. It nevertheless shows how height variation can consume the entire error allowance without changing pixel count.

Object-side telecentric optics can reduce magnification changes with height, but focus, telecentricity and calibration still require validation. The guide to telecentric measurement accuracy through depth explains why acceptable image sharpness alone is insufficient.

Calibrate at the intended measurement plane and verify performance across the usable field. A low reprojection error does not independently prove dimensional accuracy; use separate reference measurements, as discussed in calibration accuracy beyond reprojection error.

Screen product candidates without assuming qualification

Start the hardware search with area scan cameras, but filter shutter type before treating resolution as the deciding specification. The supplied catalog includes the following global-shutter candidates:

All 2 share: Sensor type CMOS, global shutter; Exposure mode Off/Once/Continuous exposure mode; Binning Supports 1 × 1, 2 × 2, 4 × 4; Decimation Supports 1 × 1, 2 × 2, 4 × 4; Lens mount C-mount; Weight Approx. 100 g (0.2 lb.); Certification CE, RoHS, KC; Mono/color Color. Values as printed in each product's data sheet; open a part for its drawing and full specifications.

The MV-CU020-90GC is a color, global-shutter camera with a GigE interface. The MV-CH120-11UC is a color, global-shutter camera with a USB3 interface.

These are screening candidates, not confirmed solutions for the stated accuracy or capture rate. Confirm native image dimensions, supported trigger modes, exposure limits and sustained acquisition at the required pixel format before ordering.

Both examples are color cameras. If the task measures monochrome edges, compare a monochrome option before committing: color-filter sampling and demosaicing can influence edge measurements. The explanation of color-camera spatial resolution covers that distinction. Sensor technology labels alone do not establish measurement accuracy.

For lens screening, the LS-FFL-0567 provides a cataloged C-mount option to investigate. It is not a confirmed match to either camera or the required field.

Use the lens finder to compare focal length, sensor coverage and mount. Confirm minimum working distance, mechanical clearance, achievable field and usable image quality before selecting a camera-and-lens combination.

Validate the complete moving measurement

Test against independently characterized references whose uncertainty is small enough for the intended acceptance decision. A calibration target used to fit the model is not, by itself, an independent validation reference.

A practical validation sequence is:

1. Establish a stationary baseline. Capture at least 30 repeated measurements without motion to estimate short-term repeatability and bias.
2. Move the reference around the field. Check the center, corners and intended measurement locations for scale and distortion effects.
3. Vary object height and camera pose. Test the full permitted mechanical range rather than only the nominal plane.
4. Introduce motion progressively. Compare stopped capture with several representative velocities, including maximum operating speed.
5. Sweep exposure. Compare the selected exposure with a shorter setting while maintaining adequate signal.
6. Exercise the full acquisition chain. Run the required frame rate, pixel format, processing and position association together.
7. Repeat across operating conditions. Include the temperature range, focus stability and mounting conditions relevant to the installation.

Record signed error against reference value, repeatability, exposure, velocity, field position and acquisition status. Thirty captures are a starting diagnostic sample, not proof of rare-event performance or long-term stability.

The relationship between small pixels and usable detail is covered in small-pixel camera selection. Use those optical checks when a higher-resolution candidate improves sampling on paper but not measured performance.

Decide when to change the method

A larger sensor or higher pixel count is not always the most practical answer. At the widest field, the required sampling may demand optics, illumination and data handling disproportionate to the task.

Consider a narrower field when only a small region contains the measurement. Consider stop-and-settle acquisition when cycle time permits. Consider multiple calibrated views when coverage matters more than simultaneous capture, while accounting for registration uncertainty and added cycle time.

If the requirement is primarily object position in motion-system coordinates, improve timestamping, pose estimation and mechanical calibration before adding pixels. If it is a dimension between two visible edges, prioritize edge quality, local scale and height sensitivity.

Before ordering, complete this checklist:

• Requirement: define accuracy, repeatability and the acceptance decision separately.
• Sampling: calculate pixel size at the widest required field.
• Motion: document relative velocity and the exposure allowance.
• Timing: verify shutter behavior, triggering and position association.
• Optics: confirm coverage, working distance, height sensitivity and mechanical fit.
• Validation: define independent references and worst-case operating tests.

Start a camera configuration once those quantities are known. For a hardware review, send the field dimensions, measurement definition, motion profile and validation criteria through application support. The useful next step is a testable imaging budget, not an unsupported accuracy claim based on megapixels.