Specular Reflection Causes Overexposure in Metal Surface Defect Detection

Why specular highlights hide defects

On a machined or polished metal surface most of the incident light leaves in a narrow lobe around the specular direction. If that lobe points into the lens, the sensor receives far more energy than the diffuse component that carries surface texture. The pixels saturate, and inside a saturated region all contrast is lost, including the contrast that a shallow scratch or a porosity pit would have produced. Reducing exposure to bring the highlight below saturation then pushes the diffuse areas into the noise floor, so a single exposure cannot cover both.

The geometry that controls it

The specular lobe follows the surface normal. For a flat part, the angle of incidence equals the angle of reflection, so moving the light source changes where the lobe lands. For a curved or worn part, the normal varies across the surface and a single fixed source will always put the lobe somewhere in the field. The practical consequence is that a defect inspection on metal almost always needs either a source with a controlled and known angular extent, or a geometry that deliberately puts the lobe outside the lens aperture.

Dark field and coaxial: opposite choices

A dark-field arrangement places the specular lobe outside the lens, so the image contains only scattered light. Surface defects scatter, so they appear bright on a dark background, which is why dark field is the standard choice for scratches and pits. Coaxial illumination places the lobe directly into the lens and produces a bright, uniform field on flat surfaces; it is the right choice when the defect changes reflectivity rather than topography, such as a stain or a coating void. Choosing between them is a decision about which physical property separates the defect from the background, and it must be made before the lens is selected.

Why dome and ring sources behave differently

A ring source illuminates from a range of azimuths at one elevation, which is useful for detecting directional defects because it produces a response that depends on scratch orientation. A dome source provides light from a wide range of elevations, which fills the specular lobe from many directions and reduces the sharp highlight but also lowers contrast for shallow topographic defects. The trade is highlight suppression against defect contrast, and it has to be judged on real parts.

What the lens contributes

A lens with a narrow acceptance cone, such as a telecentric design, sees a smaller portion of the specular lobe and therefore saturates less readily than a wide-angle lens at the same illumination. This is a secondary benefit of telecentric imaging on reflective parts, separate from its magnification stability. It does not remove the need to control geometry, but it widens the range of illumination angles that remain usable.

A procedure that converges

Start with dark field and the defects highlighted at the lowest exposure that keeps the background above the noise floor. If the defects are not visible, change the illumination azimuth before changing intensity, because azimuth changes which surface slope is highlighted while intensity only changes overall level. Only after the defect is visible should exposure and gain be tuned. Tuning exposure first is the most common reason a metal inspection plateaus below its required detection rate.

Frequently asked questions

Why does reducing exposure not reveal the defect?

Reducing exposure brings the highlight below saturation but pushes diffuse areas into the noise floor, so contrast is lost either way.

Should I use dark field or coaxial lighting on metal?

Dark field for topographic defects such as scratches and pits, coaxial when the defect changes reflectivity rather than surface shape.

Why does a ring light show scratches better in one direction?

A ring light illuminates from one elevation across many azimuths, so the response depends on scratch orientation relative to the light.

Does a dome light always improve metal inspection?

It suppresses highlights but also lowers contrast for shallow topographic defects. The trade has to be judged on real parts.

Can a telecentric lens reduce specular saturation?

It accepts a narrower cone, so it sees less of the specular lobe and saturates less readily than a wide-angle lens at the same illumination.

What should I change first when a defect is invisible?

The illumination azimuth, not the intensity. Azimuth changes which surface slope is highlighted, intensity does not.