Black O-ring machine vision lighting for reliable inspection

A black O-ring can show a dark upper band, a bright lower band and unexpected colors even when every light operates correctly. Curved rubber redirects reflections differently toward each camera, while exposure and color processing can hide the original signal. This troubleshooting workflow is for inspection stations using top and side views. You will isolate individual lights, identify clipping, compare illumination geometry and check whether the resulting images preserve the defects you actually need to detect.

1. Define the inspection before matching an appearance

Choose lighting for the defect, not for an attractive colored image. A bright band may help reveal surface texture but obscure a cut crossing that band. Uniform illumination may support contamination detection while reducing the shadows that make shallow damage visible.

Separate the inspection into observable tasks:

• Outer and inner contour: missing material, flash, shape and dimensional changes.
• Exposed surface: cuts, scratches, particles and molding marks.
• Sidewall: defects that a top camera cannot see directly.
• Full circumference: coverage across camera views, including their boundaries.

Record the smallest rejectable feature and its position on the rubber cross-section. Include acceptable molding variation so the lighting comparison does not favor a setup that highlights harmless texture.

The closest starting point is the guide to illuminating curved black rubber. The additional difficulty here is that several cameras observe different parts of a curved, closed surface. A lighting arrangement that works for the top view does not automatically work for every side view.

Keep a known acceptable part and representative defective parts throughout the tests. Do not change the inspection threshold whenever you change a light.

2. Isolate each light before diagnosing mixed colors

First stop the part and disable automatic exposure, gain and white balance. Fix the aperture, camera position and image-processing settings. Save original images rather than screenshots of a contrast-adjusted display.

Capture a baseline with the inspection lights off. Then switch on each source separately, followed by the intended combinations. For an upper green source and a lower red source, that means four captures: both off, green only, red only and both together.

Use this sequence to rank the likely causes:

Do not diagnose a wavelength problem from the displayed color alone. A rendered yellow or white patch does not identify which physical source produced it.

For a broader experimental sequence, use the lens and lighting comparison test plan. Change one variable at a time and label every saved image with the corresponding settings.

3. Test saturation with an exposure bracket

A black material can still produce a strong surface reflection. Low diffuse brightness and bright highlights can coexist within the same image.

With the part stationary and one source active, capture an exposure bracket. An illustrative sequence is 1 ms, 0.5 ms and 0.25 ms, with gain unchanged. These are test settings, not requirements for a particular camera.

For an 8-bit image, examine the individual channels for flat regions at 255. A plateau can indicate clipping, but processed output may conceal where clipping occurred earlier in the pipeline. Use linear or raw data where available and check the camera's actual output format.

Interpret the bracket as follows:

• If texture returns at shorter exposure, the original highlight was probably clipped somewhere in acquisition or processing.
• If the patch becomes darker but remains featureless, it may be a reflection with little useful defect contrast.
• If the dark band stays near the noise floor, reducing exposure alone will make that region less useful.

For a linear, unclipped signal with fixed illumination and gain, black-level-subtracted intensity should approximately follow exposure:

For example, halving exposure should approximately halve the corrected signal. Large departures warrant checking clipping, changing illumination, automatic controls or nonlinear processing.

4. Move the illumination to identify reflection geometry

An O-ring's rounded cross-section presents many surface orientations within a narrow image band. Each orientation redirects light differently. A dark strip can therefore be a geometric effect rather than a darker material region.

Move one light around the stationary part while keeping the camera fixed. Record whether the bright and dark bands move. Then repeat from another camera position. If a highlight follows the source-camera relationship, treat it as a reflection problem before changing wavelength.

Compare broad, near-axis illumination with an oblique source. Broad illumination can soften transitions. Oblique illumination can reveal surface damage through changes in scattering or shadow, but it can also make ordinary curvature dominate the image.

The guide to surface and silhouette lighting angles explains why a useful angle for surface texture may be unsuitable for contour measurement.

For top-view trials, ring lights offer illumination distributed around the viewing axis. That symmetry is useful, but it does not guarantee equal brightness across the curved rubber surface.

The LS-RNG-0426 and LS-RNG-0143 are diffused ring-light candidates for that comparison. Test their usable coverage and reflections in your actual camera geometry rather than choosing by outside diameter alone.

For side views, change upper and lower source positions independently. Opposed lights may fill a dark region, but simultaneous illumination can also remove the shadow that made a cut visible.

5. Separate spectral response from displayed color

Black rubber does not have a universal response to red, green or other illumination. Compound, surface finish, contamination and the camera's spectral response all affect the recorded signal. Determine the useful color experimentally on representative production material.

Start by comparing isolated colors at settings that avoid clipping. If the goal is to compare defect contrast rather than source brightness, make a second comparison with similar background signal levels. Record every exposure or light-output adjustment used to achieve that match.

On a color camera, inspect the individual channels as well as the rendered image. Channel sensitivity overlaps, so a green source need not appear exclusively in the green channel. White balance, a color matrix, gamma and display scaling can change the apparent hue further.

On a monochrome camera, simultaneous red and green illumination produces a combined grayscale response, not separate color labels. If the algorithm needs to distinguish the two contributions, capture them separately or use an appropriately validated color acquisition method.

A useful decision metric is defect-to-background separation, not color vividness. For selected regions, calculate:

Compare D with the variation across acceptable parts and repeated captures. A large difference in one image is not sufficient if normal texture produces similar differences elsewhere.

If glare remains dominant after geometry changes, a cross-polarization comparison can be worthwhile. It is a test option, not a guarantee that every reflection will disappear.

6. Check sampling before demanding more contrast

Lighting cannot recover a defect that the optical setup barely samples. Define the field of view and the number of pixels required across the smallest relevant feature before selecting a camera.

Consider an illustrative side view covering 30 mm horizontally. Suppose the rejectable cut is 0.10 mm wide, and the initial test target is 4 pixels across that width.

This is a sampling estimate, not a detection guarantee. Lens blur, focus, noise, defect orientation and contrast still determine whether the cut is distinguishable. Use the area scan cameras category and camera finder after establishing the required coverage and acquisition behavior.

For moving parts, include motion blur in the budget:

At an illustrative speed of 100 mm/s, limiting blur to 0.025 mm requires an exposure no longer than 0.25 ms. Whether that exposure provides enough signal must be measured with the chosen illumination.

Check coverage separately from resolution. A top view cannot inspect a hidden underside. Side views require verified overlap around the circumference; apparent overlap in mounting angles does not prove that every surface region is visible and adequately focused.

7. Choose a bench arrangement that answers a question

Use the smallest arrangement that can distinguish the competing explanations. The following comparisons are more informative than adding several sources simultaneously.

Use bar lights when you need independent control of source height, distance and direction. The LS-BAR-1276 is a diffused option. The LS-BAR-1047 is a standard option for a separate directional comparison.

These are comparison candidates, not a completed O-ring inspection system. Check emitting-area coverage, mounting clearance, electrical requirements and output at the required exposure before committing to an arrangement. Use the lighting finder to narrow the mechanical and optical options.

If separate directions consistently reveal different defects, consider sequential directional images rather than forcing one image to show everything. The guide to directional ring illumination for surface reconstruction explains the additional assumptions behind that approach. Ordinary multi-light imaging should not be treated as calibrated surface reconstruction automatically.

8. Verify the fix across parts, views and repeated captures

Once a promising setup emerges, freeze the settings and validate against the inspection requirement. Do not assess success only by whether the colored bands resemble a reference picture.

Use this checklist before ordering or finalizing the station:

• Representative parts: include acceptable texture variation, molding marks, contamination and known defects.
• Each camera view: verify focus, sampling, reflections and decision thresholds independently.
• Coverage boundaries: move a known defect through adjoining views and check for missed regions.
• Exposure margin: inspect bright areas for clipping and dark areas for unstable measurements.
• Light isolation: confirm which source contributes to each useful feature.
• Color controls: preserve the validated white balance, channel selection and processing settings.
• Timing: verify that every image corresponds to the intended source state and part position.
• Repeatability: recapture unchanged parts to distinguish noise from real surface variation.

As an initial engineering check, a set of 20 acceptable parts and 10 known defective parts can expose obvious weaknesses. Those quantities are not statistical proof of production performance. Establish the final validation sample from defect prevalence, risk and the required false-accept and false-reject limits.

Retain the original images and settings with each decision. That record makes later drift investigations much faster than comparing display screenshots.

9. Know when the usual lighting fix is wrong

More diffuse illumination is not always better. It can reduce glare yet flatten the local shading needed to detect a shallow cut. More directional illumination can improve that cut while making another orientation invisible.

Likewise, a single exposure may not cover both a very dark surface and a strong reflection. First reduce the reflection geometrically. If separate exposures or illumination states remain necessary, validate their timing and repeatability rather than assuming image combination will preserve every defect.

When dark bands persist across substantial source movement, inspect fixture obstruction, focus, aperture restrictions and lens shading. When only mixed-light colors look wrong, return to isolated-source channel data before replacing hardware.

Prevent recurrence by saving acquisition settings, recording source positions and retaining reference parts. Recheck those references after changes to the camera, lens, lighting or image-processing pipeline.

Start a lighting configuration with your field of view, smallest defect, camera views and exposure limit. For help narrowing the test arrangement, use application support and provide isolated-source images, an exposure bracket and a sketch of the viewing geometry.