Machine vision illumination types: choosing the right geometry

Poor contrast often starts with a reflection path that favors the wrong feature. More light then makes the background brighter without making the inspection easier. This comparison is for engineers choosing illumination for edges, print, scratches, recesses or reflective surfaces. Start with the feature you need to preserve, identify which reflections should reach the lens, and choose a geometry that rejects the others. The central tradeoff is simple: lighting that suppresses surface variation can also suppress the defect you need to detect.

Choose by the feature, not the light's shape

The useful distinction between illumination types is where light travels before reaching the camera. A ring describes a source arrangement; darkfield describes a reflection condition. A ring can therefore produce different image effects as its angle and distance change.

Use this matrix to select an initial geometry. The final column identifies information that the geometry may remove or make unstable.

Polarization is an optional modification to these geometries, not a replacement for choosing the reflection path. For a broader task-based starting point, use the lighting selection guide.

Understand directional light, rings and darkfield

A smooth surface reflects incident light around the mirror-reflection direction. A rough surface spreads reflected light over a wider range of directions. Real parts commonly contain both behaviors, sometimes within the same inspection region.

With directional illumination, moving the source changes which slopes appear bright and where shadows fall. This makes relief visible, but also makes the image sensitive to part rotation. A scratch aligned with the source can behave differently from an otherwise identical scratch crossing it.

Low-angle darkfield deliberately keeps the principal reflection from an undamaged surface outside the lens. Scratches, particles and edges scatter or redirect some light into the camera, appearing bright against a darker background.

Darkfield is not restricted to transparent parts. It can reveal defects on opaque surfaces when the background reflection misses the lens. Conversely, a rough opaque surface may scatter so much light that its texture overwhelms the defect.

A ring supplies light from multiple directions around the viewing axis. This can reduce directional shadows, but it does not automatically create diffuse illumination. On a polished curved part, the ring itself may appear as a bright reflection. Changing ring height changes incidence angles, coverage and the reflection pattern together.

Choose a directional source when orientation is controlled and relief is the signal. Choose a ring when several azimuths help stabilize visibility. Choose low-angle darkfield when a defect should scatter light while the acceptable surface remains dark. The scratch inspection guide explains why processing cannot recover a scratch that illumination leaves invisible.

Use a dome when reflections hide appearance

Dome lights surround the viewing region with an extended luminous environment. Instead of reflecting isolated emitters, a shiny surface reflects a broader, more continuous source. This often makes print, color boundaries and surface contamination easier to separate from glare.

The benefit depends on angular coverage, not just the outside dimensions of the housing. Part height, dome opening, viewing aperture and the gap beneath the dome all influence what the surface reflects. A large uncovered gap can remain visible as a dark band on a curved part.

The LS-DOM-0085 is a candidate for evaluating this geometry. Check its drawing against the part envelope and camera sightline before selecting it.

A dome is less attractive when the inspection depends on small shadows or changes in surface slope. Making the background smoother can also make shallow dents less visible. Use the dome lighting guide for coverage considerations rather than assuming any dome will remove every reflection.

Use coaxial light for suitable flat reflective faces

Coaxial lights introduce illumination along the viewing direction through a beamsplitter. A flat reflective face approximately perpendicular to the optical axis can return light toward the lens. Engraving, recesses and differently oriented features may then appear darker.

This is different from surrounding the lens with a ring. The illumination and observation paths substantially overlap, rather than merely occupying nearby positions.

The LS-CXL-0092 provides a coaxial candidate for a flat-face evaluation. Verify the usable illuminated field and mechanical clearance for your setup.

Coaxial illumination is not automatically glare-free. A strong return from the flat background may be exactly what creates contrast, but saturation would erase useful differences. Curved faces can show a bright central region and dark sides because their local surface normals vary.

Choose coaxial lighting when the intended background is flat and reflective, and the feature changes its reflectance or slope. Prefer a dome when substantial curvature makes a narrow range of reflection directions unreliable. If parts tilt during presentation, test the entire tilt range before treating brightness changes as defects.

Use backlighting when the outline is the answer

Backlights place the source behind the object, facing the camera. Opaque objects interrupt the light, producing silhouettes suitable for profiles, hole presence, gaps and position measurements. Transparent objects modify transmitted light through absorption, refraction and scattering, so their appearance requires separate evaluation.

Backlighting removes much of the surface-reflection problem by abandoning opaque surface information. It cannot establish whether print is correct on the camera-facing side of an opaque part.

The LS-BKL-1468 is a standard backlight candidate. The LS-BKL-0830 is a collimated alternative for comparing illumination angular spread.

Diffuse illumination sends light toward the part over a broad angular range. Collimated illumination narrows that range, which can change the apparent boundary around thick, rounded or translucent edges. Neither choice makes the imaging lens telecentric or removes perspective errors by itself.

Use the collimated versus diffuse backlight comparison when boundary position matters. Compare measured edges across the expected height range, not merely the visual sharpness of one image.

Keep the illuminated area larger than the full inspected silhouette, including position variation. Also check whether fixtures obstruct the source. A support hidden behind the part in one position can become a false contour when the part moves.

Add polarization only when it improves the signal

Cross-polarization uses a polarizer at the source and an analyzer at the lens. Their transmission axes are arranged approximately perpendicular to suppress a polarization-preserving return. Useful light that becomes partially depolarized can still reach the camera.

The outcome depends on material, coating, surface roughness, incidence angle and curvature. Metallic reflections can change the polarization state, so crossed filters should not be treated as a universal solution for polished metal. A single lens-side polarizer can reduce some reflections, but its effectiveness is also geometry-dependent.

Expect reduced light throughput. If exposure must remain short, the remaining signal may be insufficient even though glare looks better on a stationary sample.

For curved reflective features, the glare-control comparison for shiny pins provides a focused extension of this decision. Keep polarization optional until the chosen geometry has demonstrated useful feature contrast.

Work through sampling, shadows and exposure together

Consider a hypothetical inspection of a shallow raised feature on an opaque part. The horizontal field of view is 80 mm, represented by 3,200 image pixels. Object-space sampling is therefore:

A feature width of 0.10 mm spans approximately 0.10 / 0.025 = 4 pixels. That is a sampling estimate, not proof that the optics resolve it or that the measurement is accurate. Use the dimensional inspection resolution guide to separate pixel coverage from measurement capability.

Now compare directional angles. For a narrow directional source illuminating a raised feature on a flat plane, an idealized shadow estimate is:

Here the illumination angle is measured above the part plane. For a 0.10 mm height at 15 degrees, the shadow is approximately 0.10 / tan(15 degrees) = 0.37 mm, or about 15 pixels. At 45 degrees, it becomes 0.10 / tan(45 degrees) = 0.10 mm, or 4 pixels.

The lower angle amplifies relief, but also makes neighboring features more likely to cast confusing shadows. An extended source softens shadow boundaries, so this calculation is only a geometric estimate.

Finally, suppose the part moves at 500 mm/s and the permitted blur is one pixel:

Compare geometries within that exposure budget. A dome or polarized arrangement that needs a much longer exposure is not equivalent to a directional setup that meets the motion requirement. Gain can brighten recorded values, but it does not add detected photons.

Separate conflicting features instead of compromising

The usual answer is wrong when the inspection contains features that need opposite reflection conditions. A silhouette may reveal a missing edge while hiding a face mark. A dome may stabilize that mark while suppressing the shallow defect beside it.

In these cases, separate acquisitions can be more useful than a single compromise image:

• Use a backlit frame for profile and hole decisions.
• Use a diffuse front-lit frame for print or appearance.
• Use a directional or darkfield frame for relief and scattering defects.

This approach requires compatible triggering, sufficient acquisition time and reliable association between frames and parts. On moving parts, account for displacement between exposures. Registration cannot reconstruct a feature that changes visibility between views.

Do not combine sources merely to increase brightness. Adding front light to a silhouette can reduce edge contrast, while adding broad diffuse light to darkfield can raise the background. Each additional source should preserve or create a specific inspection signal.

For genuinely three-dimensional shape measurements, projected patterns and calibrated geometry may be necessary. Shadow length alone is not a general substitute for a calibrated height measurement.

Check these details before ordering

Use the lighting finder after deciding which reflection geometry fits the feature. Product dimensions alone do not establish usable coverage or inspection performance.

• Part envelope: Include position, rotation, tilt and height extremes, not just nominal placement.
• Optical clearance: Check the lens viewing cone, source opening and fixture obstruction together.
• Feature orientations: Inspect scratches or relief at several azimuths; a single favorable orientation can hide sensitivity.
• Exposure limit: Establish the motion budget before judging whether the image is bright enough.
• Brightest acceptable region: Avoid saturation where surface differences must remain measurable.
• Electrical interface: Verify supply, current, connection, dimming method and trigger requirements from the selected documentation.
• Separate acquisitions: Confirm controller channels, switching behavior and timing if multiple geometries are required.
• Acceptance samples: Include acceptable surface variation and the smallest relevant defects, rather than only obvious failures.

For a structured evaluation after selecting candidate geometries, follow the lens and lighting test plan. Keep camera processing fixed during comparisons so automatic brightness changes do not disguise unstable illumination.

Make the final choice by what remains visible

Choose the geometry that keeps the required feature distinguishable across the permitted part variation, within the available exposure time. The smoothest-looking image is not necessarily the most reliable inspection image.

Start a lighting configuration with the feature, part envelope and motion limit already defined. If several geometries remain plausible, send those constraints through application support before committing to a source arrangement.