Best Practices for Illumination in Machine Vision for Transparent Materials
Transparent components are among the most demanding objects for automated inspection. Glass, clear plastics, films, lenses and coated surfaces can appear almost invisible to a camera, while scratches, bubbles, chips, contamination and moulding marks may alter the light in subtle ways. A setup that works well for an opaque part can produce a washed-out image, bright glare or unstable defect contrast when used on clear material.
The central task is to control where light travels, how it interacts with the part and what reaches the image sensor. Good results depend on illumination geometry, wavelength, polarisation, camera settings, background control and mechanical consistency. These principles apply across packaging, automotive components, laboratory consumables, eyewear, solar products and precision optical inspection.
| Material or defect | Useful illumination approach | Main control point |
|---|---|---|
| Clear bottles and containers | Backlighting or diffuse dome lighting | Separate the silhouette from the background |
| Scratches on glass or acrylic | Low-angle dark-field lighting | Reveal scattered light from the defect |
| Surface contamination | Coaxial or diffuse frontal lighting | Reduce reflections from smooth faces |
| Internal bubbles or inclusions | Collimated backlight with controlled angles | Create contrast between the inclusion and body |
| Transparent film and sheet | Line light with uniform diffusion | Prevent banding, hotspots and wrinkles |
| Lens edges and coatings | Multi-angle lighting, often with polarisation | Capture geometry without saturating highlights |
Start with the optical geometry
The first decision is whether the inspection needs transmitted light, reflected light or a combination of both. Transmitted illumination places the part between the light source and the camera. It is usually the strongest choice for outlines, holes, missing sections, bubbles and inclusions because the clear component becomes a controlled silhouette. A high-uniformity LED backlight can turn an otherwise difficult inspection into a straightforward measurement task.
Reflected illumination is more suitable for surface scratches, scuffs, stains, coating defects and printed features. Low-angle dark-field lighting sends light across the surface so that a scratch or particle scatters light towards the camera while the surrounding area remains dark. For curved containers or lenses, several light directions may be required because a single angle can leave one side of the part invisible.
The working distance, lens field of view and angle of incidence should be designed together. A light placed too close may create local hotspots, while a light positioned too far away can lose intensity and increase sensitivity to ambient light. Transparent parts also refract light, so small changes in height or tilt can shift the apparent position of an edge. Mechanical guides and fixturing are therefore part of the optical design, not an afterthought.
Manage reflections, glare and the background
Smooth transparent surfaces behave like mirrors at particular angles. A camera may see the light source reflected directly from a bottle wall, lens face or acrylic panel, producing saturated regions that hide the actual defect. Moving the light a few degrees, changing the camera angle or using a larger diffuse source can reduce this problem. The best arrangement is usually the one that sends unwanted specular reflections away from the sensor.
Polarising filters can help where glare is caused by reflected light from non-metallic surfaces. A linear polariser on the light and a second polariser on the camera, rotated relative to the first, can suppress some reflections. This technique is effective for glass and plastics, although it can also reduce useful signal. It should be evaluated with the exact material, coating and viewing angle because stress patterns in moulded plastic may become more visible under polarised light.
Background design deserves equal attention. A bright, even background is useful for silhouette measurements, but it can hide faint internal defects. A dark background can make scattered light from scratches and contamination stand out, although it may reduce edge contrast. In many installations, a changeable background or separate image channels are preferable to forcing one lighting arrangement to perform every task.
For high-speed inspection, avoid relying on room lighting or daylight from doors and skylights. A production line near a roller door in Western Sydney or a bright factory window in Melbourne can experience large changes during the day. Enclosures, light shields and synchronised strobes provide a stable optical environment and reduce the need for aggressive software compensation.
Match wavelength, camera and exposure
Clear materials transmit some wavelengths more readily than others. Standard visible white light is a practical starting point, but red, blue, green or near-infrared illumination may provide better contrast for a particular resin, coating or contaminant. The camera’s spectral response and any protective window must be considered before selecting the light. A wavelength that looks bright to an operator may be poorly captured by the sensor.
Shorter wavelengths can emphasise fine surface scattering, while longer wavelengths may pass through certain plastics with less diffuse spread. Near-infrared can be useful when visible markings or colour variations would otherwise interfere with an internal inspection, but it will not reveal every defect. Testing several narrow-band LED options is often more productive than simply increasing white-light intensity.
Exposure should be short enough to freeze motion, especially when bottles, films or lenses move continuously. Strobing allows a high peak light output without keeping the LEDs at excessive continuous power. Gain should remain as low as practical because electronic noise can resemble small inclusions or surface contamination. Use the full useful dynamic range of the camera, while keeping specular highlights below saturation.
Telecentric lenses are valuable when the inspection measures edges, diameters or distances through a transparent object. Their low-perspective geometry reduces apparent size changes caused by part position. They are more expensive and may require a larger light source, but the repeatability can justify the investment for precision components and dimensional sorting.
Build contrast around the defect
Illumination should be selected according to the optical behaviour of the defect, rather than the general appearance of the part. A bubble inside glass may be easiest to see with collimated backlight, while a hairline scratch on the outer surface may require dark-field light from one side. A cloudy region may respond better to diffuse transmission, whereas a sharp chip on an edge can be measured from a silhouette image.
Transparent packaging often combines several inspection goals: fill level, cap presence, wall damage, contamination and label position. A single ring light may produce acceptable images for one feature and poor images for another. Separate cameras or sequential lighting states can provide cleaner data. For example, a backlight can capture the outline and a short dark-field pulse can reveal scuffs without changing the conveyor speed.
Curvature creates another challenge. A cylindrical bottle redirects light around its surface, and a flat panel may show different reflections near its edges. Diffusers, tunnel lights and multi-segment LED controllers allow intensity to be adjusted by region. Segment control is particularly useful when the top, side and base of a part have different optical behaviour.
The inspection objective should be expressed in measurable terms: minimum scratch width, smallest bubble diameter, allowable edge deviation or maximum rejected percentage. Test samples should include real defects, borderline parts and clean parts with natural variation. This prevents the system from being tuned only to attractive demonstration pieces that do not represent production conditions.
For optical products, the relationship between illumination and acceptance criteria needs careful documentation. Eyewear lenses, for example, may contain legitimate coatings, bevels and curvature that resemble damage under some lighting angles. A practical eyewear inspection case study illustrates why image quality, defect classification and production handling need to be considered as one process.
Calibrate for repeatable measurements
Calibration begins with a clean reference image and a stable reference object. Capture images with no part present to identify fixed-pattern variation, dust on the lens or uneven illumination. Then use a known target, such as a calibrated grid, edge standard or transparent reference artefact, to establish scale and verify that the camera sees the intended inspection region.
Flat-field correction can compensate for gradual brightness variation across a backlight or line-scan field. It should not be used to hide poor mechanical alignment or a damaged diffuser. Recalibration intervals depend on production conditions, but a check at the start of a shift and after maintenance is sensible for critical applications. Store the reference images with dates and equipment settings so that drift can be traced.
Focus, aperture, exposure and light intensity should be locked once validated. Automatic exposure and autofocus can produce attractive images during commissioning while causing inconsistent results in operation. A clear machine calibration guide can support a repeatable procedure, particularly when cameras, lenses or lighting modules are replaced.
Check the complete system at operating speed. Vibration, conveyor movement, compressed-air pulses and part rotation can alter the image even when a stationary test looks excellent. Record pass and fail images, monitor the brightness of a stable region and set an alert when image statistics move outside an approved range. This turns maintenance into a measurable task rather than a visual guess.
Design for Australian production conditions
Australian plants often combine long travel distances, variable ambient conditions and strong seasonal temperature changes. A line in Brisbane may deal with humidity and condensation near chilled products, while a shed in Adelaide or regional New South Wales may experience dust and high summer heat. Enclosures should be selected with suitable IP protection, ventilation and access for cleaning, especially where airborne particles can settle on diffusers and camera windows.
The local market also rewards practical integration. Many manufacturers work with system integrators that need clear specifications, readily available replacement parts and support across different states. A lighting design should document LED wavelength, working distance, diffuser type, trigger timing, connector details and expected service life. This makes it easier for a technician in Perth, Melbourne or Townsville to reproduce the setup rather than improvising with a different lamp.
Food, beverage and pharmaceutical lines may require washdown-compatible hardware, stainless-steel brackets and materials suitable for hygiene procedures. In a hot regional facility, the light output can drift as LED temperature rises, so thermal management and intensity monitoring matter. In a busy workshop, operators may describe a marginal image as “not quite right”; objective brightness limits and saved golden images translate that practical feedback into an actionable maintenance rule.
Australian daylight can be intense, and open factory layouts often allow stray light to reach inspection stations. Use shrouds, tunnel sections or enclosed cabinets around the camera and part whenever possible. Validate the system in the morning, afternoon and under production lighting, including after nearby equipment is switched on. A fair dinkum production-ready setup is one that keeps its contrast through those ordinary changes rather than working only during a controlled demonstration.
A complete illumination solution combines optical physics with disciplined production engineering. Select the geometry that makes the defect visible, suppress reflections before relying on software, choose wavelength and exposure together, and validate the result with representative samples. When the light, camera, fixture and maintenance process are designed as a single system, transparent materials become measurable products rather than unpredictable visual targets.