Is Machine Vision Right for Your Application? 5 Questions to Ask First

Machine vision can solve a remarkable range of inspection, measurement, and identification problems. But it cannot solve all of them, and there are situations where a simpler, cheaper, or completely different technology is the better answer. 

We sell machine vision components and systems for a living, so you might expect us to tell you that vision is always the right approach. We do not. In our experience, the most expensive machine vision system is one that should never have been a machine vision system in the first place. 

Before you invest in a vision project, these five questions will help you work out whether machine vision is genuinely the right fit, or whether your time and budget would be better spent elsewhere. 

1. Can a simpler technology do the job? 

Machine vision is a powerful tool. It is also a complex one. If the problem you are trying to solve can be addressed with a simpler sensor, that is almost always the better choice. Simpler means cheaper to buy, easier to install, faster to commission, and less to go wrong. 

A real example: a logistics customer needed to check whether items on the back of a lorry would exceed the height of a loading door further down the route. The system integrator was considering a 3D stereo camera to measure the maximum height of the load as the lorry drove through a checkpoint. It would have worked, but it was an expensive and complex solution to a binary question: is the load too high, yes or no? 

The answer was two through-beam sensors mounted at the critical height. If the beam breaks as the lorry passes, the load is too high. Total cost: a fraction of the 3D camera system. Reliability: higher, because there is less to go wrong. Installation time: hours instead of days. 

The lesson is not that 3D cameras are unnecessary. They are essential for many applications. The lesson is that the right question is not "can machine vision do this?" but "what is the simplest technology that solves this problem reliably?" 

2. Can you actually see what you need to inspect? 

This sounds obvious, but it catches people out more often than you would expect. Machine vision works by analysing images. If the feature you need to inspect cannot be captured in an image under any practical lighting and camera arrangement, machine vision will not help. 

Transparent objects are a common challenge. We worked on a project where a customer needed to confirm the correct punnet type was on a conveyor belt before it was filled with fruit. Multiple punnets sat across the conveyor asynchronously, so a side-mounted camera could not see them individually because they blocked each other. A top-down laser profiler would have been the obvious solution, but the punnets were clear plastic. The laser went straight through them. The material itself made the inspection optically infeasible with the available technology and budget. 

Thermal imaging through glass is another one that trips people up. Thermal cameras detect infrared radiation, and glass is opaque to the wavelengths most thermal cameras operate at. If you need to inspect a thermal seal or detect a hot spot through a glass window or screen, the thermal camera will see the glass, not the object behind it. This is a fundamental physics limitation, not a specification problem. 

The question to ask before starting any vision project is: can I get a clear image of the feature I need to inspect, under the conditions that exist on my production line? If the answer is uncertain, a feasibility study with real samples is the only way to find out before you commit to hardware. 

3. Does the environment allow it? 

A vision system that works perfectly in a controlled lab environment can fail completely on a production line. The common culprits are vibration, temperature, ambient light, and physical access. 

Vibration is particularly damaging to precision measurement applications. We evaluated a project where the customer needed sub-millimetre 3D inspection of pallets to detect cracks. A laser profiler was the technically correct approach, but the conveyor vibrated so much that the profiler could not maintain the accuracy required. A motion-compensating 3D camera existed that could have solved the problem, but the cost was disproportionate to the application. Sometimes the environment makes a technically feasible solution economically impractical. 

Thermal inspection has its own environmental constraints. If you need to inspect a thermal seal, the camera needs to be close enough to capture the heat signature before the material cools. If the production line stops frequently, parts cool before the camera can inspect them. If there are hot objects nearby, they create thermal reflections that interfere with the measurement. These are not insurmountable problems, but they need to be identified early, not discovered after the system is installed. 

If you have any doubt about whether your production environment will support a vision system, test it under real conditions before buying. Clearview's Insights Lab can help with this, but even a simple site visit with a camera on a tripod will reveal more than a desk-based specification exercise. 

4. Can you change the process to make vision easier? 

This is the question that most people do not think to ask, and it often delivers the biggest cost saving. 

A practical example: consider a scan tunnel that reads labels on boxes. If the labels can appear on any of five sides, you need cameras covering all five sides. That is five cameras, five lighting setups, five mounting positions, and the software to manage all of them. If you can change the loading process so that operators place boxes with the label facing the same way every time, you need one camera instead of five. The cost difference is significant, and the system is simpler, faster, and easier to maintain. 

This is not always possible. Some production lines cannot be modified, and some products genuinely need inspection from multiple angles. But before you specify a complex multi-camera system, ask whether a simple change to the upstream process could reduce the vision system to something much simpler. The cheapest camera is the one you do not need to install. 

5. Does the ROI justify the investment? 

Machine vision is an investment, not an expense, but like any investment, it needs to deliver a return. The business case for vision typically rests on one or more of the following: reducing manual inspection labour, catching defects that manual inspection misses, meeting regulatory or retailer compliance requirements, reducing waste and rework, or preventing costly product recalls. 

For many applications, the payback period is measured in months. Label verification systems, for example, pay for themselves the first time they prevent a product withdrawal that would have cost tens or hundreds of thousands of pounds. In-line dimensional measurement pays back through reduced scrap rates and fewer downstream assembly failures. 

Where the ROI is harder to justify is when the defect rate is very low, the production volume is small, or the cost of a missed defect is negligible. If you are inspecting 50 parts a day and a rejected part costs you a few pence, a machine vision system that costs several thousand pounds may never pay for itself. That is an honest assessment, and we would rather tell you that upfront than sell you a system you do not need. 

For help building a business case, our Budget Planning Guide includes a framework for calculating ROI.

The honest answer 

Machine vision is the right answer for a large number of inspection, measurement, and identification problems. It is proven, reliable, and cost-effective when properly specified. 

It is not the right answer for everything. If a simpler sensor will do the job, use the simpler sensor. If the environment will not support it, address the environment first. If the process can be changed to make the inspection easier, change the process. And if the ROI does not stack up, do not force it. 

If you are unsure whether machine vision is the right approach for your application, the most useful thing you can do is talk to someone who will give you an honest answer. Clearview's engineering team has been doing this for 18 years, and we have told plenty of customers that vision was not the right solution for their specific situation. We would rather lose a sale than sell you something that does not work. 

Get in touch: info@clearview-imaging.com | +44 (0)1844 217270