Types of machine vision lighting: ring, backlight, coaxial, dome and dark field
A practical guide to machine vision lighting geometries — ring, bar, backlight, coaxial, dome and dark field — explaining what each reveals, the surfaces it suits, and how to choose the right one directly from your inspection problem.
Why geometry, not brightness, is the real variable
The most common mistake in machine vision lighting is treating it as a brightness problem. An engineer adds a brighter light, the image washes out or the defect stays invisible, and the camera or algorithm gets the blame. In reality, the variable that decides whether a feature is visible is the angle at which light strikes the part relative to the camera's line of sight. Light that reflects from a surface into the lens makes that region bright; light that reflects away from the lens leaves it dark. Every lighting geometry is simply a way of engineering where those bright and dark zones fall.
This is governed by a single physical distinction. On a matte (diffuse) surface, incoming light scatters in many directions, so the surface looks broadly similar from a wide range of viewing angles — these surfaces are forgiving and tolerate many lighting choices. On a specular (glossy, mirror-like) surface, light obeys the law of reflection: the angle of incidence equals the angle of reflection. A specular surface behaves like a mirror, so you do not so much illuminate the part as arrange for a controlled reflection to reach the camera. Most real parts sit between these extremes, and the same part can behave diffusely in one region and specularly in another.
Because of this, the goal of lighting design is contrast: making the feature of interest appear consistently brighter or darker than its surroundings, part after part. A well-lit image often lets a simple threshold or blob tool succeed where a poorly lit one would demand fragile, compute-heavy processing. Getting the optics right at the front end is almost always cheaper and more reliable than compensating in software downstream.
Front-light geometries: ring, bar and spot
Front lighting illuminates the part from the same side as the camera and is the workhorse of general inspection. A ring light mounts around the lens and throws light onto the part from a moderate angle, fairly close to the optical axis. It is compact, easy to integrate, and well suited to presence/absence checks, dimensional gauging of matte parts, and general feature location. Its weakness is specular surfaces: on anything shiny a ring light tends to produce a bright reflected ring or hotspot, and on curved glossy parts a distracting circular glare. Ring lights come in various working distances and beam angles, and low-angle variants shade toward dark-field behaviour.
Bar lights (also called linear lights) are rectangular and deliver directional, often grazing illumination. Used in pairs or arrays they cover large fields and light line-scan webs, and — when angled low across a surface — they raise the contrast of texture, scratches, and embossed or engraved marks by casting shadows. Their directionality is the point: you can place a bar to skim across a feature so its raised or recessed edges catch the light. They are a flexible building block precisely because position and angle are entirely under your control.
Spot lights provide a concentrated beam for small targets, for back-illuminating fibre bundles, or as the source feeding a coaxial or telecentric arrangement; a true collimated beam usually requires added collimating optics rather than the bare LED. As a class, front lights share one trade-off: they are simple and bright but expose the surface's own reflectivity directly to the camera, which is exactly what you want on matte parts and exactly what you want to avoid on shiny ones.
- Ring light: best for matte parts, presence/absence, general gauging; prone to glare on shiny or curved surfaces.
- Bar / linear light: directional and grazing; ideal for texture, scratches and embossing via shadowing, and for line-scan webs.
- Spot light: concentrated beam for small features or as a feed for coaxial/telecentric optics.
- Low-angle ring: a hybrid that pushes toward dark field, emphasising surface relief and edges.
Backlight: silhouettes, edges and through-features
A backlight sits behind the part, facing the camera, so the part appears as a dark silhouette against a bright, uniform field. This is the highest-contrast, most repeatable geometry available and the natural choice whenever the inspection depends on an outline or a gap. Backlights excel at precise dimensional measurement, edge detection, counting, finding missing or broken features, and revealing holes, cracks or contamination in translucent materials where light leaks through. Because the result is a near-binary bright/dark image, edge-finding and gauging tools can reach sub-pixel repeatability with minimal tuning.
The limitation is equally clear: a backlight reports the part's silhouette and its translucency but almost nothing about its top surface. You cannot read a printed label, find a surface scratch, or judge surface finish with a backlight. There must also be physical room behind the part and a way to present it reliably between the light and the camera — on conveyors this often means inspecting at a gap, over a glass plate, or against the open frame of the line.
For the most demanding gauging, a collimated backlight (often sold as a telecentric backlight) pairs with a telecentric lens. Collimated light travels in near-parallel rays, which sharpens edges and reduces the edge-position error that diffuse backlights cause on parts with appreciable thickness, where light wraps slightly around the rim. This combination is standard for high-accuracy measurement of turned parts, connector pins and similar precision components, where edge ambiguity of even a pixel translates into measurement error.
Coaxial and dome: taming specular and curved surfaces
Coaxial (on-axis) lighting injects light through a beamsplitter so the illumination travels along the same axis as the camera's view. Light strikes the part close to perpendicular and, on a flat specular surface, reflects straight back into the lens. Flat, mirror-like regions then appear uniformly bright, while anything that tilts the surface — an engraved mark, a scratch, a depression, a solder feature — reflects light away and appears dark. Coaxial light is the classic choice for flat reflective surfaces: reading laser-etched marks on metal or silicon, and inspecting mirrors, wafers, polished plates and PCB features. It is unforgiving of tilt, so it suits flat parts presented squarely, not curved ones.
A dome light (also called a cloudy-day or diffuse-dome illuminator) solves the opposite problem: highly reflective, curved or irregular surfaces. The part sits under a hemispherical diffuser lit from the rim, so light arrives from nearly every angle at once. This suppresses hotspots and glare, producing a flat, even, near shadow-free image even on crumpled foil, chrome, glossy curved plastic, or rounded metal. Dome lighting is the standard for reading print on cans and bottles, inspecting reflective curved housings, and any case where you must mask the surface's mirror behaviour and read colour or print uniformly. The trade-offs are physical size, the camera aperture needed at the top, and the fact that a dome deliberately removes the directional shadows you would actually want when hunting for fine surface defects.
Between these sits the flat-dome, or coaxial-diffuse, light: a low-profile unit that combines a diffuse component with an on-axis component in a compact housing. It is a practical compromise when you need dome-like glare suppression but lack the height for a full hemisphere — common in tight machine enclosures and on flat or gently curved reflective parts.
Dark field and low-angle: scratches, edges and relief
Dark-field (low-angle) lighting illuminates the part at a very shallow grazing angle, nearly parallel to the surface. On a smooth surface this light skims across and reflects away from the camera, so the background stays dark. But wherever the surface is interrupted — a scratch, an engraved or embossed character, a particle, an edge, a raised defect — that feature scatters the grazing light back into the lens and lights up brightly against the dark surroundings. It inverts bright-field thinking: you make the defects glow rather than the surface.
This geometry is the specialist's tool for surface and edge inspection. It is exceptional at finding fine scratches, surface particles, lint and contamination, and at reading embossed, debossed or engraved characters where a flat front light would see almost no contrast. A low-angle ring delivers this from all directions at once; individual bar lights at a grazing angle deliver it directionally, which helps when defect orientation matters. The cost is that dark field is sensitive to working distance and angle, lights a limited field uniformly, and can be noisy on rough or textured surfaces whose natural relief lights up alongside the real defects.
Dark field also pairs well with multi-shot techniques. Capturing several images under light from different directions — the basis of photometric stereo — and combining them lets you separate true topographic defects from printed or coloured features, because real surface relief shifts its shading with the light direction while flat print does not. This is increasingly common for demanding surface inspection where a single exposure cannot tell a defect apart from harmless surface decoration.
A decision framework: from surface and feature to geometry
The reliable way to choose is to characterise the problem along two axes — the surface and the feature — and let those point to a geometry. First classify the surface: matte (diffuse), specular (shiny), translucent, or textured. Then classify the target feature: a silhouette or edge, a printed mark, surface relief such as a scratch or embossed character, or a colour difference. The intersection of those two answers narrows the field quickly, and a third factor — whether the part is flat, curved or moving — usually settles the final choice.
A few mappings cover most real jobs. A silhouette or dimension on any surface points to a backlight, collimated if accuracy is critical. A printed mark or colour on a shiny curved part points to a dome; on a flat reflective part, to coaxial or a flat-dome. Surface relief — scratches, particles, embossing — points to dark field or a grazing bar. General features on a matte part are well served by a ring or bar. When the feature is a colour contrast, geometry gets you an even image and the wavelength of the light then does the separating work.
Two practical cautions. First, lighting is never an isolated choice: the geometry sets the requirements for the lens, the bandpass filter, the controller and the trigger, and on moving lines it usually forces strobed, hardware-triggered illumination so a brief, intense pulse can freeze motion and overpower ambient light. Second, when no single geometry cleanly separates the defect from the background, do not force one — prototype with real good and bad parts, judge the raw image contrast before any software, and be ready to combine geometries or use multi-shot acquisition. Lighting is cheap to trial relative to the cost of a misdiagnosed inspection, and the cleaner the image, the simpler and more robust everything downstream becomes.
- Silhouette / dimension → backlight (collimated + telecentric lens for high accuracy).
- Print or colour on a shiny curved part → dome; on a flat reflective part → coaxial or flat-dome.
- Scratches, particles, embossing → dark field or a grazing bar light.
- General features on matte parts → ring or bar light.
- Colour-based contrast → even geometry plus a chosen wavelength (and a matching bandpass filter) to do the separating.
Frequently asked questions
Related categories
- Machine vision lighting
- Backlights
- Coaxial / axial
- Dome lights
- Dark field
- Ring lights
Related guides
How to choose machine vision lighting
A practical guide to choosing machine vision lighting: geometry, wavelength, strobe vs continuous, ambient control and how to maximise feature contrast for reliable inspection.
How to choose a lens for machine vision
A practical engineer's guide to selecting machine-vision lenses: focal length from FOV and working distance, sensor format, resolving power, aperture, depth of field and telecentricity.
Monochrome vs color cameras
When to choose monochrome vs color machine-vision cameras: Bayer light loss and effective resolution, color-dependent tasks, wavelength-specific lighting, NIR sensitivity and processing cost.