Skip to main content

Machine vision image (data) creation

or the way of light

For a long time, I struggled to truly understand something one of the early vision pioneers once told me:

"You don’t see the reality; you only see the reflected light."

If you think about it long enough, it gets spooky. Is what we see really there? With cameras, at least, we can prove it: we don’t actually see objects, we see the light that reflects off them.

camera system with special illumination checking dimensional accuracy

Why reflected light defines information in every vision system

Light = Information

Take the colour black, for example. Technically, black is simply the absence of light. Polar bears, for instance, have black skin, not because we can see it, but because it absorbs light instead of reflecting it. That energy gets converted into warmth.

It’s a powerful reminder: light is the key to everything in machine vision. To extract meaningful information from an image, you must shape the illumination so that the photons you need reach the sensor, while the ones you don’t are kept out

How often do we really think about this?

Honestly. When setting up a vision system, do you truly ask yourself: "How am I shaping the light? What do I want to reflect and what do I want to keep out?" Or have you ever been surprised by an image that looked fine, but didn’t actually contain the data you needed?

Conclusion: Follow the path of light

In the end, it all comes down to this. Understanding and deliberately shaping the path of light
is the foundation for reliable, meaningful image data.

Key lighting pitfalls that distort your data

  • Too diffuse or uncontrolled lighting
  • Reflections and glare
  • Shadows and absorption

How consciously do we shape light for reliable image creation in machine vision?

  • Choose the right lighting type: direct, diffuse, dome, ring ... each has its place depending on the surface.
  • Pay attention to angles: the angle of incidence affects where light goes ... cameras only see what bounces back.
  • Understand materials: matte vs. glossy, transparent vs. opaque ... they all affect how light behaves.
  • Don't forget sensor and optics setup: exposure, aperture, filters ... without good lighting, they can't help you.
industrial inspection showing a PCB board inspected with a camera and a dome light

Why every vision set-up should start with illumination

Next time you set up a camera system, don't start with the sensor specs or the software. Start with the light.

Ask yourself:
What is not being reflected into the image?

Because at the end of the day, we don't capture reality, we capture reflected light.

Be honest, did you really make this clear when working with cameras?
When was the last time the light surprised you?

Mail to Patrick