Automating visual inspections is an “impossible game” and a “crap game.”
ovitAs I mentioned during the seminar at the Image Sensing Expo, I’ve been in this industry for many years, and there’s something I’ve been feeling more and more strongly lately.
That is,
Automating visual inspections is, by its very nature, an “impossible game” and a “crap game.”
That is to say.
Of course, this is not meant to disparage engineers.
Actually, it's the opposite.
Even when a group of talented engineers comes together, they struggle not because they lack technical skills, but because the topic itself is difficult.
Barrier 1: The Cost Barrier
Visual inspection is a necessary step in quality assurance.
However, in many cases,
This is not a process that directly adds value to the product itself.
Therefore, from a company's perspective,
- Areas for Cost Reduction
- A process I’d like to keep as inexpensive as possible
will be.
On the other hand, when trying to automate,
- Equipment Costs
- Development Costs
- Adjustment Fee
- Maintenance Costs
- Operating expenses
will occur.
In other words,
The costs we can afford and the costs we need to incur don't align.
Here is the first obstacle.
Barrier #2: The Barrier of Diversity
The scope of visual inspections is truly diverse.
- There are many varieties
- The workpiece shapes are different
- The materials are different
- The colors are different
- The lighting conditions are different
And above all,
Each item is unique.
A method that worked at one site can rarely be applied as-is at another site.
As a result,
- The Accumulation of Individual Optimal Solutions
- Knowledge Dependence on Individuals
- Cannot expand horizontally
That's how the situation will be.
In the software industry, it's like starting from scratch every single time.
Barrier ③: The Barrier of Subjectivity
What makes it even more troublesome is this wall.
Visual inspections are inherently subjective.
For example,
- Is this scratch okay?
- Is this stain a no-no?
- Is this color variation within acceptable limits?
Such judgments vary from person to person.
In other words,
- The criteria are vague
- Opinions vary from person to person.
- The goal itself cannot be fixed
There is a problem, namely...
It is not a world with clear-cut right answers, like shogi or chess.
To begin with, the correct answer keeps changing.
That's why it's an "impossible game."
The costs are high.
The subjects vary.
Even the correct answer is ambiguous.
When I think about it calmly,
There's no way this can be automated that easily.
The reason why automating visual inspections is difficult is that,
It’s not because image processing is difficult,
nor is it because AI is still in its infancy.
The topic itself is difficult.
That's why I,
I believe that automating visual inspections is structurally an "impossible task."
And then, “crap game”
What's even more difficult is that,
The return on investment tends to be low relative to the effort involved.
It's different every time.
Adjustments are needed every time.
Every time, the know-how gets reset.
This will burn out the engineers.
That's why I sometimes,
People say that automating visual inspections is a "crap game."
Of course, I say this with love (lol).
Reasons to Do It Anyway
So, why do I keep doing it?
The reason is simple.
It's because the people on the ground are struggling.
The labor shortage is accelerating.
The number of experienced inspectors is also declining.
And quality requirements are becoming increasingly stringent.
In other words,
Automating visual inspections is difficult.
But I have to do it.
Where is the path to victory?
Lately, I've been,
Rather than increasing the number of individual cases,
Turning Achievements into Assets
I believe that is important.
- Don't start from scratch every time
- Share know-how
- Accumulate data on high-quality products
- Creating a System That Drives Continuous Improvement
And so,
In a world where every item is unique, we gradually increase the common elements.
That's the only way to win.
Automating visual inspections is definitely an “impossible game” and a “crap game.”
But that's exactly why it's interesting.
And there is still room for improvement.
We will continue to,
We will continue to strive to develop visual inspection methods that are truly practical in the field.
