Machine vision system selection by application requirements

An unfamiliar system name tells you little about whether its replacement will inspect your parts reliably. It might describe an integrated camera, an image-processing controller, or a complete station with dedicated optics and lighting. For maintenance teams and machine builders, the useful starting point is therefore the inspection requirement, not the nameplate. You can turn that requirement into a measurable specification, shortlist suitable components, and define an acceptance test without assuming that a historical product identity determines the right architecture.

Start with the inspection decision, not the equipment name

Machine vision system selection starts by defining what the machine must decide. Finding a part, measuring its diameter, reading a code, and detecting a surface scratch require different image information. A system that succeeds at one task may fail at another despite using the same camera resolution.

Write a requirements sheet before comparing equipment. Include the following:

• Decision: presence, count, position, dimension, appearance, identification, or a combination.
• Smallest relevant feature: the smallest defect or marking that changes the decision.
• Inspection area: the complete region that must remain visible, including positioning variation.
• Motion: speed during exposure, possible stops, and variation in part spacing.
• Appearance variation: acceptable differences in finish, color, orientation, and presentation.
• Output: the required result, destination, deadline, and behavior when inspection fails.
• Acceptance criteria: allowable false rejects, missed defects, measurement error, and invalid inspections.

Distinguish a defect's visible size from the required measurement tolerance. Detecting a dark mark measuring 0.30 mm across is not equivalent to measuring an edge position within 0.30 mm. Contrast, calibration, focus, and the decision method affect those tasks differently.

The system components buyer's guide explains the roles of acquisition, illumination, processing, and communication. Use those functions to describe the existing station, even when its original documentation is incomplete.

If historical identification remains necessary for maintenance, use existing records and legitimate manufacturer or distributor documentation. Do not treat a similar abbreviation as evidence of electrical, software, or mechanical compatibility.

Calculate sampling, blur, and the decision deadline

Three calculations eliminate many unsuitable configurations before you compare individual parts: spatial sampling, exposure time, and available response time. They describe different limits and should remain separate in the specification.

Spatial sampling

For a first estimate, divide the required field width by the smallest feature size, then multiply by your chosen sampling target.

Suppose a part occupies 180 mm horizontally. Allowing 10 mm of positioning margin on each side gives a 200 mm field width. Assume the inspection must detect a 0.30 mm feature, with an initial target of 6 pixels across it.

The arithmetic is:

• Required horizontal pixels = 200 × 6 / 0.30 = 4,000 pixels.
• Object-space pixel size = 200 / 4,000 = 0.05 mm per pixel.
• For a 150 mm field height, required vertical pixels = 150 / 0.05 = 3,000 pixels.
• Total sampling = 4,000 × 3,000 = 12 million pixels.

Six pixels per feature is a trial assumption, not a universal detection guarantee. Validate it using the lowest acceptable contrast and representative defects. The camera resolution and measurement accuracy guide explains why pixel count alone does not establish measurement performance.

Motion blur

If the part travels at 500 mm/s and your initial blur allowance is 0.5 pixel, the permitted movement during exposure is 0.025 mm.

This sets a light-and-exposure requirement. It does not establish that a particular camera and light will produce adequate contrast at that exposure.

Decision deadline

If the reject point is 150 mm downstream, the travel time is 150 / 500 = 0.30 s, or 300 ms. Acquisition, transfer, processing, communication, and actuator response must all fit inside that interval with margin.

At a 100 mm part pitch, the arrival rate is 500 / 100 = 5 parts per second. This throughput requirement is separate from the deadline for each individual result. The conveyor camera selection guide develops these motion-related tradeoffs further.

Choose the camera by image requirements and motion

Start with area scan cameras when one image can contain the required inspection area. Use the camera finder to narrow resolution, shutter type, interface, and color response after calculating your requirements.

Global shutters expose the image pixels over a common interval. Rolling shutters expose different rows at different times, which can distort moving objects. A short exposure reduces blur within each row but does not necessarily remove rolling-shutter geometry errors.

Color is useful when the decision depends on color differences. It is not automatically beneficial for dimensional inspection or monochrome markings. Compare the actual feature contrast under controlled illumination before accepting the additional color-processing requirements.

The LS-CMR-0166 is a 12 MP USB3 camera with a rolling shutter and a printed resolution of 4000 × 3000. The MV-CU020-90GC is a 2.5 MP GigE color camera with a global shutter and C-mount.

All 2 share: Sensor size 1/1.7". Values as printed in each product's data sheet; open a part for its drawing and full specifications.

These examples illustrate a tradeoff, not interchangeable replacements. Matching the worked example's pixel count does not establish suitability for continuous motion. Conversely, having a global shutter does not establish sufficient spatial sampling.

Before accepting either architecture, verify:

• Full-resolution operation with the required exposure and trigger mode.
• Available image dimensions, not just the total megapixel figure.
• Trigger-to-exposure timing and sustained acquisition behavior.
• Interface support in the intended processing environment.
• Lens mount, active sensor dimensions, and accessible configuration controls.

Treat any missing specification as an open requirement, not permission to infer it from a similar product.

Select optics around field coverage and feature contrast

The lens must cover the sensor and resolve the relevant feature across the usable field. Matching the thread alone is insufficient. Begin with machine vision lenses, then use the lens finder to narrow mount, image circle, focal length, and application type.

Field of view depends on sensor dimensions, magnification, and camera position. Working distance also has to leave practical clearance for the part and lighting. Confirm how the lens documentation defines working distance before using its value in a mechanical layout.

A nominal sensor-format label is a useful screening filter, but active sensor dimensions and the lens image circle provide the more direct coverage check. Lens performance must also suit the required object detail, especially toward the field edges.

The LS-FFL-0602 is an 8.5 mm C-mount lens specified for a 1" sensor format.

The LS-FFL-0567 is a 16 mm C-mount lens specified for a 1.1" sensor format.

These are focal-length candidates for layout evaluation, not a validated pairing with every camera above. Do not choose between them without the camera's active sensor dimensions and the available working distance.

Test focus at the center, corners, and expected part-height extremes. Stopping down can increase depth of field, but it also reduces collected light and can introduce diffraction-related detail loss. The lens selection guide provides the broader optical checks.

If dimensional results change when the part height changes, investigate perspective sensitivity. A conventional lens may be unsuitable even when its image looks sharp.

Make lighting reveal the decision feature

Lighting should separate acceptable and defective conditions, not simply make the image bright. Start by examining how the relevant surface reflects light and whether the feature is a silhouette, printed contrast, raised edge, or surface disturbance.

Bar lights allow directional illumination and flexible placement. They are useful candidates when changing the incidence direction reveals edges or texture. Diffusion can soften visible source structure, but it does not guarantee freedom from glare.

The LS-BAR-1047 is a standard bar light with a printed size of 290 x 30 mm. The LS-BAR-1276 is a diffused white bar light with a printed size of 372 x 28 mm.

Values as printed in each product's data sheet; open a part for its drawing and full specifications.

Neither size nor the word “diffused” establishes illumination coverage at your inspection plane. Measure the image response at the proposed distance and angle. Use the lighting finder after those trials identify the required geometry and color.

For the worked exposure limit, test whether the feature remains visible without clipping bright regions or hiding dark defects in noise. If pulsed operation is necessary, verify the light and controller's documented compatibility rather than assuming a continuous light accepts overdrive.

The lighting selection guide helps distinguish surface, silhouette, and glare-control approaches. For a combined trial procedure, use the lens and lighting test plan.

Specify processing and controller responsibilities

A camera interface is not a complete integration specification. Decide which device receives triggers, acquires images, runs inspection, associates results with parts, and communicates with the machine controller.

Use the standalone controller selection guide to evaluate dedicated processing architecture. Specify required image inputs and machine communication before choosing a controller.

For the worked 300 ms deadline, an illustrative allocation might be 40 ms for acquisition and transfer, 60 ms for processing, 20 ms for communication, and 30 ms for actuator response. The sum is 150 ms, leaving 150 ms for variation and margin. These are planning allocations, not promised component performance.

Measure the complete path under sustained load. The image-based control latency guide explains why camera frame rate alone cannot establish response time.

Define behavior for missing triggers, failed inspections, stale images, communication loss, and restart. A result should identify the correct part, not merely report the latest completed calculation. Ordinary inspection outputs must not replace a required safety-rated function.

Compare candidates using evidence, not feature totals

Create a selection matrix with one row per mandatory requirement. For each candidate, record the supporting documentation or the test that demonstrates compliance. Keep desirable features separate from rejection criteria.

A useful scoring order is:

1. Image feasibility: the required feature remains distinguishable across acceptable part variation.
2. Timing feasibility: acquisition and decisions meet the sustained rate and individual deadline.
3. Integration feasibility: triggering, communication, mounting, and software support fit the machine.
4. Maintainability: settings can be backed up, restored, and checked after replacement.
5. Commercial fit: the complete component set and support requirements suit the project.

Do not let optional tools compensate for a failed mandatory requirement. A sophisticated algorithm cannot restore a feature that was never resolved or was hidden by glare.

For each unresolved item, assign a next action: obtain a documented specification, capture a test image, measure timing, or revise the requirement. This produces a defensible shortlist without pretending that incomplete catalog data proves system compatibility.

Validate with representative parts before ordering

Build a test set containing acceptable variation as well as clear failures. Include borderline examples near the acceptance limit, because obvious defects rarely reveal the real selection risk.

As an illustrative trial, collect 100 acceptable parts, 30 defective parts, and 20 borderline examples. Those quantities can expose basic weaknesses, but they do not establish a low production escape rate. The final validation sample must match the risk and required statistical confidence.

Capture images at the minimum and maximum expected part heights, positions, speeds, and surface variations. Freeze exposure, gain, focus, aperture, and lighting settings while comparing candidates. Automatic adjustments can otherwise conceal a weak optical setup.

Check image quality before tuning thresholds. Record the failure mechanism: insufficient contrast, blur, clipped highlights, missing coverage, incorrect geometry, or processing error. Changing several variables together makes the cause harder to identify.

Run the complete result path, including reject timing and part association. Test restart and communication recovery with the machine in a safe commissioning state. Save settings and acceptance images so replacement hardware can be checked against the same reference.

Know when a conventional setup is the wrong answer

A single area-scan image is not always the best starting point. Reconsider the architecture when the inspection depends on hidden surfaces, substantial height variation, uncontrolled overlap, or continuous material beyond a practical frame.

Similarly, adding resolution is the wrong response when glare hides the defect. Increasing frame rate is the wrong response when results are associated with the wrong parts. A different lens cannot correct unstable part presentation by itself.

Separate these failure classes before expanding the component list. First improve presentation or optical contrast, then recalculate sampling and timing. Consider another acquisition geometry only when the task requires information that the current view cannot supply.

For a replacement project, functional equivalence matters more than copying every historical component. Preserve the necessary inspection behavior, interfaces, and acceptance criteria; reconsider implementation details that no longer serve those requirements.

Final selection checklist and next step

Before approving the shortlist, confirm that:

• The inspection decision and acceptance limits are written down.
• Sampling covers the smallest relevant feature across the full field.
• Exposure and shutter behavior suit actual motion.
• The lens covers the sensor and passes corner and height tests.
• Lighting reveals the feature across acceptable surface variation.
• Processing and communication meet a measured end-to-end deadline.
• Failure behavior, settings backup, and replacement verification are defined.
• Remaining assumptions have an assigned test or documentation request.

Start the camera configuration process with your field size, feature size, motion, and required output. If an essential compatibility detail is missing, send the requirements sheet through application support before committing to a component combination.