Artificial Intelligence is everywhere – from your phone’s camera to the coffee machine that knows when to clean itself, even to the cars that recognize pedestrians.
But when it comes to industrial inspection, things get a bit more serious. You don’t just need to know what you’re looking at – you need to know how big, how precise, and how consistent it is.
And while AI makes for great headlines, in industrial image processing, it’s not the full story.
At Eye Vision Technology, we see AI as an incredibly powerful tool – but one that works best in combination with classical, rule-based image processing. Let’s be honest: AI can recognize, classify, and segment like never before, but when it comes to precise measurement and inspection, the classical algorithms still hold the crown.
AI shines at recognition – classical tools rule at precision
Take a simple example: measuring a silver disk.
Using classical tools, it’s easy to find and measure – as long as the lighting is stable and the object looks exactly as expected. AI, however, can locate that same disk even if it’s rotated or slightly shaded, thanks to deep learning and pre-trained networks.
But let’s say you now have two similar round objects that differ only slightly in texture. AI can instantly tell them apart, while classical image processing may struggle. Yet, if you then need to measure their diameters or distances, the rule-based algorithms deliver the accuracy you need.
That’s why at EVT we say: Don’t choose between AI and classical vision – combine them.
When reflection confuses the camera
We all know those shiny metal surfaces that make inspection engineers sigh. Imagine trying to find a fish-shaped imprint on a reflective background. Classical tools often fail here, but AI can handle it easily – it learns to “see” beyond reflections and lighting variations.
But what if you then need to measure that fish precisely?
AI can tell you “Yes, it’s a fish,” but not how long or wide it is. Classical measurement tools finish the job.
The smart solution? Let AI do the recognition, then hand over to rule-based metrology for precise evaluation.
AI + 3D Vision + Robotics: A Powerful Trio
Now picture a robot sorting Ricola candies – yes, that’s a real example we trained.
The AI identifies the candy type; the 3D vision captures depth and position; and the robot knows exactly where and how to grip. None of these systems alone could solve the task – but together, they deliver fast, reliable results.
Or think of bin picking – for years, one of the toughest automation challenges. Even with 3D cameras, separating overlapping parts was difficult. Now, with AI-driven segmentation and point cloud analysis, the robot knows not only what to pick, but also exactly where and how.
Scratches, defects, and anomalies – let AI do the boring work
Detecting scratches or surface anomalies is another case where AI shines. While classical tools need carefully tuned thresholds, AI can learn what “normal” looks like and flag any deviations – even subtle ones. Especially in variable environments, that flexibility is priceless.
Not everything needs AI
There are still many tasks where AI simply doesn’t make sense.
Precise geometric measurements, edge detection, or calibration often rely on predictable, rule-based tools. And that’s perfectly fine – in fact, it’s necessary.
A license plate reader, on the other hand? That’s AI’s playground. Recognition tasks like that are where AI outperforms any classical algorithm.
Conclusion: Choose what solves the problem, not what’s trending
AI is not replacing classical image processing – it’s enhancing it. The best results come from knowing when to use each approach. That’s why Eye Vision Technology offers both: tools that recognize like AI and measure like a metrologist.
Because in the end, what matters isn’t the hype – it’s the solution.
Want to see it in action?
If you’re facing a challenge in inspection, measurement, or automation, talk to us.
Our EyeVision Software combines AI, Deep Learning, and classical image processing in one flexible platform – so you can focus on your task, not on choosing between algorithms.
