Machine vision system components: a buyer's guide

A successful inspection starts with a visible, repeatable difference between acceptable and unacceptable parts—not a camera purchase. First-time buyers need to connect part presentation, lighting, optics, capture timing, processing and machine communication into one testable system. This guide gives you an ordering sequence, practical sizing calculations and a bench-trial checklist. Use it to identify which specifications matter before committing to components that look suitable individually but do not work together.

1. Define the inspection decision before choosing hardware

Write the inspection requirement as a decision with a measurable acceptance criterion. “Inspect the assembly” is too broad. “Detect a missing retaining clip anywhere within the permitted assembly position” identifies both the feature and the variation you must accommodate.

Record these inputs before opening a component finder:

• The acceptable condition and each known failure condition.
• The smallest feature that changes the decision.
• The permitted part position, rotation and height variation.
• The production rate and available decision time.
• The consequences of accepting a defective part or rejecting a good one.
• Whether you need presence detection, identification, measurement or position guidance.

A presence check and a dimensional measurement can require very different optics, even on the same assembly. For measurement, distinguish image sampling from demonstrated measurement uncertainty. More pixels alone do not establish accuracy.

The machine vision versus computer vision guide explains the wider distinction between interpreting images and delivering a repeatable production decision.

2. Treat part presentation as a system component

Presentation determines what the camera can see and how much variation the software must handle. A fixture, mechanical guide or controlled transport path may improve reliability more than a higher-resolution camera.

Specify position variation numerically. If a part can shift by 5 mm in either direction, allow for that movement on both sides of the inspection area. Also check rotation: a rectangular part can occupy more image width when rotated than when aligned.

Height variation matters separately. It can change focus, magnification and the apparent position of edges. A flat fixture does not help if the inspected surface itself changes height between assemblies.

Choose between two capture conditions:

Do not assume every task needs a software search. A well-constrained presence check may use a fixed image region. Position-varying parts usually need localization before the inspection tools run.

3. Create contrast, then select the light controller

Lighting should make the relevant difference easier to separate from normal part variation. Start with the feature, not with the brightest source available.

For an outside profile, test illumination behind the part. For surface marks, compare illumination directions. For reflective surfaces, change the reflection geometry or diffusion before increasing exposure. A bright image can still contain almost no useful defect contrast.

During a bench trial, keep exposure and gain fixed while comparing lighting arrangements. Otherwise, automatic adjustments can conceal differences between setups. Repeat the comparison with surrounding lights switched between their expected operating states.

Light controllers regulate or switch compatible lights, but they cannot correct an unsuitable lighting angle. Use the light controller finder after identifying the required electrical output, channel count and operating mode.

The LS-CTL-0070 is a digital controller with a catalog-listed strobe feature. That feature alone does not establish compatibility with a particular light or capture sequence.

Confirm the light’s drive requirements, controller output, trigger interface and permitted operating conditions together. For pulsed operation, the strobe timing and duty-cycle guide explains the checks that matter before applying overdrive.

4. Size the field of view around real variation

Field of view is the physical area recorded by the camera. It must contain the inspection region throughout the allowed part movement, not just during an ideal setup photograph.

Consider a hypothetical part measuring 100 mm wide and 60 mm high. Allow 5 mm of position variation on each side, plus another 5 mm of visible margin beyond that variation:

With a sensor aspect ratio of 4:3, a 120 mm image width corresponds to 90 mm of image height. The part fits within that view, with additional vertical space. A different sensor aspect ratio could require a different width or height to contain the same inspection envelope.

Reserve room for locating features and calibration marks when they must appear in the production image. Do not add an arbitrary oversized view: every increase reduces the number of pixels covering a given feature.

The calibration target coverage and margin guide helps distinguish useful calibration coverage from unnecessary image area.

5. Calculate camera sampling and the motion budget

Suppose the smallest relevant feature is 0.25 mm wide. As an initial bench-test target, allocate 5 pixels across it. This is a starting assumption, not a universal detection guarantee; contrast, shape and processing affect the actual requirement.

This calculation establishes a horizontal sampling requirement. It does not prove that a particular camera and lens will resolve the feature. Optical blur, focus errors, reflections and image processing can remove useful detail before the decision stage.

If the part moves at 500 mm/s, calculate the exposure permitted by an illustrative 1-pixel blur budget:

This estimates blur from motion during exposure. It does not account for rolling-shutter distortion or trigger-position variation. A short exposure may also demand more illumination than the original lighting trial provided.

Use the moving-part exposure guide to connect exposure limits with lighting and motion. For measurement tasks, the dimensional inspection resolution guide explains why sampling and accuracy need separate budgets.

6. Choose the camera around capture conditions

Once you know field of view, sampling and exposure requirements, compare area scan cameras. Use the camera finder to narrow the interface, shutter type, sensor format and required pixel dimensions.

The most useful first comparison is not simply total megapixels:

The LS-CMR-0053 uses GigE, while the LS-CMR-0166 uses USB3. Both have catalog-listed rolling shutters, so neither should be treated as an automatic choice for moving-part inspection.

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 are comparison candidates, not a validated system recommendation. Confirm trigger support, exposure behavior, software compatibility and image transfer performance from the current product documentation. The industrial camera specifications guide provides a deeper checklist for that review.

7. Match the lens to the sensor and working distance

Select machine vision lenses using sensor coverage, mount, field of view and working distance together. Nominal focal length alone is not enough.

For an initial thin-lens estimate, suppose a hypothetical sensor has an active width of 8.8 mm. With the earlier field width and an approximate object distance of 250 mm:

This is a screening estimate. Practical working distance is commonly specified from a lens reference surface, while the thin-lens calculation uses a principal plane. Real lens construction and focusing behavior change the installed geometry.

The LS-FFL-0567 is a C-mount lens with a catalog-listed 16 mm focal length. Treat it as a candidate for review, not proof that the illustrative geometry will work.

Use the lens finder and lens selection configurator to narrow candidates. Then verify focus range, sensor coverage, mechanical clearance and edge image quality. The lens focal-length selection guide explains how to move from an estimate to an installed choice.

8. Plan processing and communication as one timing chain

The processing stage must do more than run an inspection algorithm. It must associate the correct image with the correct part, detect acquisition faults and deliver an actionable decision before the machine needs it.

Record at least these events during development:

1. Part-detection or capture request.
2. Exposure start and image availability.
3. Inspection completion.
4. Decision transmission.
5. Machine acknowledgment or reject action.

At 120 parts/minute, the average interval is 60 / 120 = 0.5 seconds per part. That does not automatically provide 0.5 seconds for processing. If a decision is needed after only 0.3 seconds of travel, that shorter interval governs the deadline.

Separate throughput from latency. A pipelined system may process several parts concurrently while each individual decision takes longer than the interval between parts. It still needs reliable part identifiers and sufficient buffering.

Define explicit states for acceptable, rejected, no image, processing error and communication timeout. Do not silently classify a missing image as an acceptable part.

For hardware-triggered capture, review the trigger-source and line-configuration checks. For PLC data exchange, the vision communication guide explains result mapping and handshakes. Ordinary inspection outputs are not a substitute for a safety-rated control function.

9. Validate the full chain in a bench trial

A bench trial should challenge the complete imaging and decision chain, rather than demonstrate one successful photograph. Include acceptable parts near their permitted limits and defects near the rejection boundary.

For example, begin with 30 acceptable parts and 30 defective parts spanning the known variation. Capture each under 3 allowed presentation conditions, producing 60 × 3 = 180 images. This is an exploratory dataset, not proof of a production failure rate.

Keep separate samples for tuning and evaluation. Repeated captures of the same part help assess repeatability, but they do not replace physical variation between parts.

Use a grouped trial checklist:

• Presentation: Test extreme positions, rotations and surface heights; these reveal coverage and focus failures.
• Lighting: Repeat with permitted ambient changes; this exposes dependence on uncontrolled illumination.
• Image quality: Inspect the smallest features at the center and edges; this checks usable detail across the view.
• Decision quality: Record false accepts and false rejects separately; they carry different operational consequences.
• Timing: Measure individual deadlines and sustained throughput; average processing time can hide occasional overruns.
• Fault handling: Test missing captures and interrupted communication; the machine must recognize an invalid inspection.
• Restart behavior: Reload saved settings and repeat known samples; startup should restore the validated configuration.

Save raw images, component settings and decision records together. That evidence makes later changes traceable.

10. Check the ordering list—and know when to change approach

Before ordering, reconcile the camera mount, lens coverage, focusing range, light drive requirements, controller connections and acquisition interface. Include cables, power supplies, mounting hardware, triggers and software licensing in the bill of materials.

Do not assume similar connector shapes establish electrical compatibility. Confirm pin assignments and signal requirements in the supplied documentation. Likewise, a controller’s power rating does not establish its minimum pulse duration or suitability for overdrive.

The usual fixed-camera arrangement may be wrong when:

• The required surface is hidden from the selected viewpoint.
• Uncontrolled height changes invalidate a planar measurement.
• A continuous web or long moving product calls for a different acquisition method.
• A tiny feature requires so much sampling that multiple views become more practical.
• Part variation produces more contrast change than the defect itself.

In those cases, revisit presentation, viewpoints or inspection method before buying more pixels or adding processing complexity.

Start the camera-led system configurator once the bench trial establishes the requirements. If compatibility remains uncertain, send the field of view, smallest feature, motion conditions and representative images through applications support before finalizing the component list.