AI Machine Vision for Food Manufacturing Quality Inspection
Wednesday 13 May 2026 · 3:45
This is due to the natural variation in colour, consistency and shape of the products they’re dealing with.
In this case study, we show how AI-powered machine vision systems were used to drive efficiency on the production line - whilst also helping to save marriages all over the UK…
The Food Manufacturing Customer
Our customer is a well-known ready meal manufacturer in the UK, in this case, manufacturing fish pies.
If you’ve ever eaten a fish pie, you’ll know that one of the variables is where the fish ends up being positioned within the pie.
Usually, the fish is evenly distributed around the pie, but sometimes you can end up with all the fish on one side only.
The most common negative feedback the manufacturer received was about the distribution of fish within the pie.
The Fish Pie Quality Inspection Challenge
A machine vision system was used for two parts of the inspection process.
- The primary application was to detect the quantity of fish in the pie
- The second part of the process determined whether the fish was evenly distributed
Salmon is added to the pie first. Being pink in colour, detecting the salmon inside a paper tray is straightforward; we could use traditional tools for this part of the process.
A 2D, rules-based colour camera doing a simple pixel count would be able to tell us where the salmon was located within the tray.
However, when it came to detecting cod in the fish pie, things became more complicated. This is because the cod is added to the pie after the creamy sauce goes in.
Being a very similar colour to the sauce, the cod was more difficult to locate. Additionally, the colour of the cod and sauce varies day-to-day; some days the cod is the same colour as the sauce was yesterday, and vice versa.
One of the big challenges for machine vision and automation applications within the food manufacturing industry is that we’re dealing with organic products.
One day the cod might be a light creamy colour, and the next it may be slightly darker. Similarly, the sauce could vary, sometimes matching the colour of the cod.
What the system would recognise as cod, it might think is sauce the following day.
So the question is, how do we get machine vision systems to pick out products that vary in colour and consistency so much?
The AI Machine Vision Solution
For this food manufacturing machine vision application, there was only one piece of kit we’d turn to; the Cognex In-Sight D902C powered by Cognex ViDi software.
The Cognex In-Sight D902C is a high-resolution, deep learning-powered colour machine vision system. It’s perfect for complex, automated inspection applications like this, where colour detection is also critical.
The Cognex In-Sight ViDi software is a deep learning-based image analysis software that combines AI capabilities with the In-Sight spreadsheet software.
Deep Learning, like all AI tools, requires repetition to teach the system what it should be looking for.
In this case, the process was to draw around the cod in many different images. The system could then learn to detect the cod, while ignoring the salmon or creamy sauce.
Accurate Food Inspection
Using this system, trained properly by our machine vision experts, our customer was able to reliably pick out the cod inside the creamy sauce and check that it was distributed properly.
In fact, the system became so good, it was as accurate as a human would have been at checking the product.
So next time you’re tucking into a fish pie, you can be assured everyone will be getting an equal quantity of fish, thanks to our machine vision expertise!
Bringing Efficiency to Your Food Manufacturing
If your food manufacturing production line could benefit from efficiency savings with machine vision, but you’re unsure where to start, book a call with us now, and we’ll be happy to talk you through some options.
