Skip to main content

Why feasibility studies are essential.
Even after 27 years of experience :-)

I have been working in industrial machine vision for 27 years. During this time, I have designed, analyzed, and calculated more than 3,000 systems, about one third of them finally made it into series production.

You might think that after so many projects, you know exactly how to approach a vision task and how to solve it. And yes, experience helps enormously.

Still,
or maybe precisely because of that:
Experience does not protect you from mistakes!

Patrick Gailer
sample setup for feasibility testing
setup illumination tests

Even today, I make one or two mistakes in every single system design. I oversimplify things, overlook a parameter, or neglect an effect that later comes back to bite me on the shop floor.

Typical candidates are:

  • Depth of field
  • Exposure margins
  • Contrast that looks different in reality than expected
  • Noise or shading
  • Timings that suddenly turn out to be too slow in the real process

You name it...

This is not a sign of incompetence. It is the reality of complex technical systems.

Yes. Some things are simple.

Of course, there are applications that can be implemented “just like that” with the right level of experience.
If the contrast is right, many tasks work reliably, such as object matching, code reading, or OCR.

But as soon as it gets complex, you have to test

Once the contrast is not clearly defined, or the application becomes more complex, you have to try it out.

That is exactly what feasibility studies are for.

I am repeatedly surprised how an image actually looks like when the selected components are used together.
Which defects or features have far less contrast than expected. And which effects appear, that were not on anyone’s radar during planning.

Then it becomes necessary to test whether the problem can be solved robustly in software, or whether the hardware needs to be adapted.

You do not gain these insights from datasheets. You only get them from real images.

Fixing problems later is extremly expensive

This is not just our experience. It is also well documented.

A well-known NASA study on "Error Cost Escalation Through the Project Life Cycle" shows that the cost of correcting errors can increase by up to a factor of 50, if problems have to be fixed during production, instead of during the evaluation phase.

What only requires some time and testing during the feasibility phase can later lead to weeks of rework, mechanical modifications, and production downtime.

test laboratory setup

Robustness  is not created at the desk

A feasibility study must not be limited to “it somehow works”. Robustness has to be high enough to cover not only object variations, but also production-related variations such as changing light conditions, contamination, mechanical tolerances, etc.

Feasibility studies are neither a luxury nor a sales argument.
They are an essential part of professional machine vision.

And if someone like me, with 27 years of experience and thousands of projects done, still tests every new application, it is not because I am unsure.
It is because I know, how expensive it is, not to test.

What is your experience with feasibility studies in machine vision? Have early tests saved your project, or have missing evaluations caused problems later on? Let us test for you!