Knowledge · Lighting
A practical buyer’s guide to industrial light sources, ring, bar, backlight, dome and strobe, and how to match geometry, intensity and timing to matte, shiny, transparent and DPM surfaces.
Written for engineers, OEMs and purchasing teams specifying their first vision station, calm guidance, not catalogue fiction.
In Computer Vision, the sensor only records what the light reveals. A high-resolution camera with a carefully chosen lens still fails when glare washes out edges, when shadows hide a scratch, or when a transparent blister looks identical to empty air. Lighting is not a decorative accessory after the BOM is frozen; it is the primary tool that creates contrast between defect and background, code modules and metal, fill level and bottle wall.
project teams often reverse the order: pick megapixels, then optics, then “add some LEDs.” Experienced stations reverse that again. They ask what must be visible, which surface physics will fight them, how fast the part moves, and only then which sensor and light geometry can freeze a stable image. Software can threshold, filter and classify, but it cannot invent photons that never reached the pixel. If two grey levels are identical in the raw image, no algorithm will reliably separate them shift after shift.
That is why lighting projects so often outrank camera brand debates in real commissioning. Changing incidence angle by a few degrees, switching from continuous wash to a short strobe, or replacing a flat ring with a dome can turn an intermittent false reject into a quiet, stable recipe. The same principle applies whether you run a dedicated smart camera, a PC-based stack, or an identification reader: the optical chain starts at the light source. Sedeco’s catalogue groups these products under the Computer Vision lighting hub, use that page as the product entry point while this article explains the decision logic.
In one sentence
Computer Vision lighting is the engineered placement of intensity, angle, colour and time so that the features you care about become high-contrast, repeatable image structure, before software ever runs.
Industrial light sources are defined less by wattage marketing and more by geometry: where photons come from relative to the camera axis and the part surface. The five families below cover most first systems. Combinations are normal, a backlight for silhouette plus a low-angle bar for surface texture, or a dome for DPM with a strobe controller for conveyor freeze.
A ring mounts around the lens and floods the field from near-coaxial directions. It is the default “front light” for many presence checks, label reads and matte assemblies because installation is compact and the illuminated zone tracks the FOV. Rings win when the surface is reasonably diffuse and you want even illumination without a large external fixture. They struggle on highly specular parts: the camera often sits inside the reflection cone, so hot spots bloom across critical edges. Low-angle or dark-field rings (LEDs aimed more sideways) help reveal scratches and embossing by lighting the texture while the flat field stays darker, useful for DPM and surface defects, less ideal for flat printed contrast.
Bar (or linear) lights are rectangular LED arrays you aim like a studio softbox. You control incidence angle freely: high angle for bright-field fill, grazing angle for dark-field texture, dual bars for wrapping a cylinder. Bars win on long parts, conveyor lanes, and any station where you need asymmetric light that a ring cannot provide. Two opposing bars reduce shadow from tall components. For line-oriented inspection they pair naturally with motion; for area scan they remain the most flexible external geometry in a machine builder’s toolkit. Watch working distance and beam uniformity, cheap bars can show intensity falloff that your algorithm mistakes for a defect gradient.
A backlight sits behind the part and creates a bright field against which the silhouette appears dark. It is the classic choice for gauging outer contours, hole presence, fill-level menisci (with care), and many transparent or translucent objects where front light creates confusing caustics. Backlights win when the feature of interest is a boundary in space, not a surface colour. They fail when you need surface print, scratch or colour, the silhouette cannot show what is painted on the face. Collimated or telecentric backlights improve edge crispness for precision measurement; diffuse panels are simpler and often enough for presence and coarse gauging on area scan stations.
A dome (or cloudy-day illuminator) surrounds the part with diffuse light from many directions so specular reflections return soft, even grey instead of hard glare. Domes win on shiny metals, blister packs, glossy plastics and many DPM marks where a ring would produce uncontrollable hotspots. The trade-off is bulk, access for robots or conveyors, and sometimes reduced shadow contrast that you actually wanted for 3D-like cues. Open-bottom or partial domes help when the part must enter from below. If a dome is mechanically impossible, polarized coaxial light or multi-angle bars are the usual substitutes, not “just a brighter ring.”
Strobe is less a shape than a timing mode: short, intense pulses synchronized to exposure. Many ring, bar and backlight families are strobe-ready when driven by a proper controller. Strobe wins on moving conveyors, vibrating feeders and any takt where continuous light would demand long exposures and motion blur. It also reduces ambient plant-light competition because your pulse can outshine windows and overhead fluorescents during the microsecond window that matters. Continuous mode still wins for static fixtures, slow indexing, and setups where flicker or EMC constraints complicate pulsing, but if parts move, plan strobe early.
| Geometry | Wins when | Watch-outs |
|---|---|---|
| Ring | Compact FOV, matte parts, integrated with camera | Specular hot spots; limited angle control |
| Bar | Custom angles, long parts, dark-field texture | Mounting space; uniformity along length |
| Backlight | Silhouette, holes, many transparent outlines | No surface detail; access behind part |
| Dome | Shiny / curved / DPM glare control | Bulk; softens wanted shadows |
| Strobe | Motion freeze, ambient rejection | Sync, controller, peak current design |
Browse current families and form factors on the lighting hub. This guide does not invent Sedeco SKU ratings, confirm intensity, wavelength and strobe specs on the product pages for the models you shortlist.
Continuous light is on whenever the station is powered. It is simple to wire, easy to debug with the naked eye, and sufficient when parts are stationary or crawl slowly enough that a short camera exposure still gathers photons without blur. The risk on fast lines is that you lengthen exposure to “get more light,” which smears edges and quietly destroys gauging and code modules. Raising sensor gain instead adds noise. The healthier pattern is more photons in less time, which is exactly what strobe provides.
Strobe lighting fires a bright pulse that overlaps the camera’s exposure window. Typical industrial practice: the PLC or sensor triggers acquisition; the camera asserts a strobe-out or the light controller accepts a shared trigger; the pulse width is set shorter than or equal to the exposure, with enough intensity that the histogram sits in a healthy mid-range without clipping specular peaks. Global-shutter area scan cameras pair especially well with strobe because the whole frame samples the same instant. Rolling shutter plus continuous light on a moving web is a classic source of geometric distortion.
Architecture choice still matters: a smart device with integrated LEDs may handle strobe internally, while a PC-based cell with external bars needs an explicit controller and cable plan. Settle that context with smart vs PC-based vision before you freeze the lighting BOM.
The same ring light that flatters cardboard can destroy a brushed stainless Data Matrix. Surface class should drive geometry before brand preference.
Paper labels, rubber, unfinished plastics and many painted parts scatter light widely. Bright-field front lighting, ring or high-angle bars, usually delivers even grey fields and readable print. Watch for varnish or film overlaminates that secretly turn a “matte” label into a partial mirror; a slight angle change or polarizer often restores stability. Colour LED choice (red, white, blue, IR) can lift contrast on tinted inks without moving to a colour camera.
Machined metal, foil, glossy mouldings and wet parts reflect the light source like a mirror. If the camera sees the LED reflection, modules and scratches disappear into white. Domes, polarized coaxial setups, and carefully aimed bars that keep the specular lobe away from the lens are the standard cures. Never “fix” glare by overexposing and hoping software clips it, you also clip the signal you needed. Test oily and dry samples; oil films change the reflection story overnight.
Glass, PET, blister film and clear vials transmit, refract and reflect at once. Backlights excel for outline and fill silhouette; front light can show print on the face or particles in the volume depending on angle. Beware double images from front and back surfaces. Collimation, telecentric optics and narrow-band LEDs sometimes reduce confusing caustics. For liquid levels, lighting trials with min/max fill and foam are worth more than datasheet promises.
Castings, fabrics, embossing and orange-peel coatings create micro-shadows. Low-angle dark-field lighting exaggerates texture, helpful for dents and missing material, harmful when you only wanted printed contrast on a rough label. Choose consciously: bright-field for print, dark-field for topography. Mixed recipes (two lights, two exposures, or dual-channel setups) appear in advanced cells but add integration cost; start with one dominant geometry when possible.
Laser etch, dot peen and electrochemical marks often have low albedo contrast: the “dark” modules are depth or scatter, not ink. Domes, low-angle rings and multi-zone lighting are common winners; what works on brushed stainless may fail on sandblasted aluminium. Treat DPM as a lighting project first, decode package second. Bring worst-case production parts to any trial �� polished golden samples lie. Pair this article with identification guidance when codes are the KPI, and keep the lighting hub open while you shortlist dome and dark-field options.
| Surface | Typical light | Buyer hint |
|---|---|---|
| Matte | Ring / high-angle bar | Check for hidden gloss films |
| Shiny | Dome / polarized / angled bars | Trial oily + dry samples |
| Transparent | Backlight ± careful front | Separate silhouette vs print goals |
| Textured | Dark-field low angle | Do not confuse texture with defect |
| DPM | Dome / low-angle / multi-zone | Lighting trial before SKU lock |
Polarizers filter light by orientation. A common machine-vision pattern places a polarizer on the light and a crossed analyzer on the lens so specular glare (which preserves polarization) is attenuated while diffuse scatter from the feature still reaches the sensor. That is why polarized rings and coaxial lights appear so often on glossy packs and coated metals.
Polarization is not magic: it reduces intensity, so you may need brighter LEDs or strobe; some plastics scramble polarization; and extreme angles can still leak glare. Use it as a targeted tool after geometry is approximately right, not as a substitute for choosing dome versus backlight. When you evaluate options on the lighting product pages, note whether polarizing films or kits are available for the form factor you need.
Rule of thumb
If moving the light a few degrees removes the hotspot, fix geometry first. If every angle still mirrors the LED into the lens, try dome or crossed polarization next.
Datasheets list luminous area, wavelength, IP rating and strobe current. prioritize the list below. Do not invent Sedeco part numbers or intensity claims from this article, confirm current options on the hub and product sheets.
The light must cover the FOV with acceptable uniformity at your mechanical WD. Undersized lights create centre-bright images; oversized lights waste space and may illuminate neighbouring stations. Match to lens FOV plans from the lenses discussion.
Ring, bar, backlight or dome, and whether the robot, gripper or conveyor physically allows that shape. Cable exits, brackets and vibration locking are part of the spec, not afterthoughts.
Red is common for mono sensors; white for colour inspection; IR for some print-through or ambient-hiding cases; blue/UV for fluorescence or specialty contrast. Pick wavelength for material contrast, not aesthetics.
If the line moves, verify strobe mode, maximum pulse current, recommended controller and duty cycle at your parts-per-minute. Continuous-only lights on fast conveyors push you toward blur or gain, both expensive later.
Trigger inputs, intensity control (analogue, PWM, digital), and safe wiring to the camera strobe line determine how cleanly the station commissions. Integrated lights on smart devices simplify this; external bars on PC cells need explicit design.
IP rating, temperature, washdown chemicals and ESD context matter as much as optical niceties. A fogged dome window is an inspection outage.
High throughput, label and film contrast, frequent SKU changeovers. Strobe plus global shutter keeps edges crisp; polarizers tame glossy overlaminates. Colour or wavelength swaps help weak ink on tinted film without redesigning the whole cell.
DPM, oily finishes and mixed alloys dominate. Dome and low-angle lighting trials are non-negotiable. Traceability stations often need image evidence, so lighting recipes must be reproducible across shifts and plants.
Tiny features, reflective solder and ESD-aware mounts. Compact rings and coaxial lights are common; IR or specialized wavelengths sometimes separate materials. Pair lighting design with the camera FOV early, small modules leave little margin for glare.
Transparent primary packs, serialization codes and validation documentation. Backlights for silhouette, controlled front light for print, and locked recipes for auditors. Reproducibility beats peak brightness.
Often yes for first failures. Extra megapixels cannot create contrast that geometry never produced. Start with surface physics and light, then size the sensor.
When shiny or curved surfaces throw the ring’s reflection into the image. Domes diffuse that glare; rings stay better for compact matte FOVs.
If motion during exposure blurs your smallest feature, yes, or slow the line. Continuous light only works when exposure can stay short enough without starving the sensor.
Sometimes with multi-angle or multi-colour banks and recipe changes. Mixed matte and mirror SKUs may need two geometries. Design for the hardest part, not the average.
FOV and WD from the lens define how large the illuminated field must be; the camera shutter and trigger define how you sync strobe. Treat them as one optical chain.
Collect worst-case samples, state the feature and takt, then run a short lighting trial. Begin from Knowledge, the feasibility mindset, and the lighting hub.
You now have the buyer model: lighting creates contrast before software runs; geometry must match surface class; strobe and exposure must share one timing story; and product selection belongs on Sedeco’s lighting hub, not in invented SKU lists. Pair lights with the right camera and lens, prove the recipe on real parts, then lock mechanics so the image stays boringly stable.
Ring, bar, backlight, dome and strobe families for your cell.
Match shutter, resolution and I/O to the lighting plan.
Sample-based trials before you freeze the BOM.
Send parts, takt and FOV, we help select light geometry.
Related: Kennis / Knowledge, lenses, area scan cameras, smart vs PC vision.