Choosing Light Wavelength, Polarisation and Filters

The purpose of lighting is not to brighten the image but to separate the feature being judged from everything around it in grey level. The same scratch is nearly invisible under one setup and obvious under another, and the difference is not brightness but contrast. Once that is understood, lighting stops being a matter of adding two more lamps and hoping and becomes directional: decide which two things must be told apart, then find the variable that amplifies the difference between them. Wavelength, polarisation and filtering are the three variables that change contrast directly.

Colour contrast first, brightness second

A surface appears a given colour because it reflects that wavelength and absorbs the complementary one. That yields a practical rule: illuminate a part with its own colour and it brightens; illuminate it with the complementary colour and it darkens. A concrete case: inspecting a yellow mark on white packaging. Under white light both are bright and the grey difference is small, making the threshold hard to place. Switch to blue light and the yellow absorbs it and darkens while the white keeps reflecting and stays bright, so the separation opens up and the decision becomes easy. The reverse case matters too: for dark characters on a yellow label, blue light is the wrong choice, because darkening the yellow brings it closer to the characters. Yellow or red light is what that task wants.

Monochrome light has another benefit that is easy to undervalue: it collapses the colour dimension. White light contains every wavelength, so the reflections of different materials average together and the grey difference flattens out. Under monochrome light each material returns a grey level set by its reflectance at that wavelength, so the differences stretch, and batch to batch colour variation within one material is suppressed. This is a large part of why a monochrome camera with monochrome lighting is more stable than a colour camera for most inspection tasks.

Polarisation: the tool for glare

Specular glare from machined metal, plastic parts, glass and laminated surfaces blows out parts of the image and drowns defects that were there all along. Polarisation works by placing a polarising filter in front of the light source so the illumination reaching the part has a single polarisation direction, then placing a second filter crossed against it in front of the lens. Specular reflection largely preserves the polarisation of the incident light and is blocked by the lens filter, while light that entered the material and scattered back out has its polarisation scrambled and passes through. The result is suppressed glare and clearer texture and defects beneath the surface.

The cost has to be accounted for in advance. A crossed pair of polarisers removes most of the light energy, often leaving only a few tenths of the original throughput, which means either a stronger light or a longer exposure. For a part that is genuinely reflective, that money is usually better spent than repeated algorithm tuning. For a matt surface that already scatters diffusely, polarisation is pure loss with no return. The test is simple: turn off the ambient light, shine a strong torch at the part, and if there is a blinding highlight, polarisation has value.

Filters: keeping ambient light out of the lens

Light on a shop floor keeps changing: sunlight slants in when a roller door opens, a forklift drags a moving reflection across the workstation, a row of ceiling lights fails, the next cell installs new equipment. None of that light has anything to do with the inspection, yet all of it changes the grey levels in the image. When a monochrome source is used, adding a narrow band filter of the same wavelength in front of the lens keeps the great majority of ambient light away from the sensor. This is the cheapest image stability measure available; one filter costs far less than the time later spent chasing those fluctuations.

A filter has a precondition: the source and the filter must share a centre wavelength, and the source must be narrow band. A white lamp behind a narrow band filter simply loses most of its energy and the image only gets darker; an 850 nm infrared source behind an 850 nm filter is an effective combination. Narrow band filters are also angle sensitive — the centre wavelength shifts as the angle of incidence increases — so filtering is weaker at the edges of a wide field than at the centre. Ambient light is only one of the environmental influences on imaging; dust, oil mist, temperature and vibration all play a part, and they are covered together in shop floor environment.

A three-step order of selection