MicrosoftTeams image 32

Machine Visionas a solution to the shortage of skilled workers?

Solutions for medium-sized businesses

Germany as an industrial location is being severely tested due to many different crises. The pandemic years with the resulting supply chain bottlenecks caused delays and dislocations in the procurement market, rising energy prices are causing increased costs in production, and the resulting inflation is raising concerns of a severe recession for the coming year 2023.

And it feels like all of this together is the tip of the iceberg seen so far that the large and small industrial ships are currently heading towards. Other dangers still lie hidden under water – you can already sense them, but you can’t really see them yet.

  • Intensification of political conflicts with associated even more difficult procurement situation
  • Severe cuts due to extreme situations in the climate
  • Rising shortage of skilled workers

The latter issue is perhaps the most tangible at the moment and at the same time already felt. The Ifo Institute just released a report on August 2 describing that 44% of the manufacturing sector is having trouble filling jobs. In addition, current political developments make it increasingly conceivable that more production will be brought back to Germany as an industrial location, which will also further exacerbate the situation.

SMEs should automate more

There must be an even stronger rethinking of the industry, especially among SMEs, and the step into digitalization as well as automation must be considered even more strongly. The fact is that companies that are already having difficulty filling positions will have an even greater problem with skilled workers in the coming years – even in spite of the recession. The problem should therefore not just be shunted off to HR departments.

Especially in medium-sized companies, a lot of skilled personnel is still employed for activities that could be replaced by technical solutions – such as Machine Vision. It is important to mention at this point, we are not talking in this article about replacing and saving a specialist – those days are fortunately over. We talk about the more sensible use of a skilled worker in the company whose tasks are simplified by digitalized and automated solutions, and about harnessing the potential of the worker when simple tasks can be taken over by technology.

SMEs – like corporations – need to automate their production processes more now so that companies are positioned for when the skills shortage really hits – when the steamship can no longer avoid the iceberg. The illustration is a bit drastic, but I wanted to finish the top image. And yet, the difficulty of attracting skilled workers increases in proportion to the need for automation for their own production in the market. If one thinks now of the current topics of renewable energy, such as the construction of photovoltaic systems on one’s own home and the associated delivery times and cost explosion, one can quickly guess what will happen when the time comes:

Long delivery times, combined with high costs, with declining in-house production.

What technical solutions are available for medium-sized companies in the field of Machine Vision?

BV systems can provide a great remedy, they are the inspecting and decisive eye and thus an important part of a process chain during production:

Example 1: Quality check

Whether at goods receipt, in the value chain and at the end of the production line. BV systems can follow the process chain and carry out checks at defined positions and document them. Whether it is surfaces or contours, whether measurements or target/actual comparisons are to be set up, BV systems work 24h without fatigue effects.

Example 2: Completeness check

When assembling a component, it is possible to check precisely whether all subcomponents are correctly installed and whether components are missing or incorrectly positioned for safety reasons.

Example 3: Track & Trace systems

Tracking components during the production chain is a topic that is becoming increasingly popular. Be it via fingerprints or applied coding. With additional plant data, all process steps can be restored here at the end and conclusions drawn about defective parts.

Machine Vision is still a black box

As is often the case, the solutions are already very close and are actually already provided by many providers. But why is there still such a great need to catch up in the field of Machine Vision, especially for SMEs? The main reasons for this are:

  • High acquisition costs
  • Further dependencies on the supplier even after delivery
  • Difficult to operate and high training requirements
  • Poor acceptance – employees’ concerns
  • Unmanageable system and high complexity

All of this rightly discourages SMBs from moving to new systems. Even if you consider that especially at the beginning there is still some need for adjustment for the system itself. Machine Vision is still a black box in some ways – an image goes in the front, a result comes out the back. It is not directly clear how the decision was made and whether it is correct according to the specifications. This is a deterrent.

Ready-made BV solutions already provide a good remedy

So it is just right that there is a trend towards solution-oriented BV systems. Certain tasks are already solved and offered in advance by the manufacturer. Many tasks, such as label recognition, text recognition, contour recognition and many more, are already represented in the portfolio of major manufacturers. Here you can really only advertise to take the time as a company to look at the individual solutions of the different providers. Also, companies are already offering AI solutions with online creation of training sets, which is a step in the right direction. However, many of these solutions help only up to a certain degree of the task. For everything else, project planning with all its advantages and disadvantages is needed again. This means that the high acquisition costs and the dependence on the supplier remain. The technology is therefore only interesting for those who drive high volumes.

Machine Vision must also be democratized!

The trend toward democratization of robots is already evident in robotics. Behind this is the understanding that robots must be easy and simple to program. Why? Simply so that the robots can be applied to several different tasks. If a medium-sized company initially needs the robot for task A, it can be reprogrammed for task B a few months later – completely independently of the supplier. The lifetime of the product becomes longer, profitability increases.

And the state of democratization of Machine Vision? Currently, at least according to my feeling so “well”:

Here we are just talking about more specialized customer-specific solutions, not standard solutions, where the problems mentioned above are particularly common. Unfortunately, reprogramming an Machine Visionfrom task A to task B is not so easy here, because too many different influences have to be taken into account. Questions like:

  • Which camera system consisting of hardware, optics and light is the right one?
  • What is the cycle time that must be maintained with the processing task?
  • What endpoint information is needed?

are actually still the most tangible and yet already difficult to answer. Especially in image processing, the non-visible factors and issues play a special role in solving the problem:

  • What data throughput and image size must my code be able to handle?
  • Which processing strategy do I want to use?
  • Do I need a conventional or an AI-based solution approach?

These are issues that make it difficult for small and medium-sized businesses in particular to find independent solutions in their day-to-day business. And yet, the step into self-solution is often dared. Recognition rates up to 85% are indeed possible and at the same time a good first iterative step, but they are still useless for the production world. Every further step costs money, time and nerves at the same time.

A kind of partial democratization has therefore already been achieved, but it is unfortunately not sufficient for production.

But here it must and can be mentioned that many large providers have the ambition to go exactly in this direction. Provide applications and solutions that are usable by the layperson, not just the expert. If we look ahead over the next few years, a lot will happen in this segment of Machine Vision in particular. Be it through good and easy to understand software or online teaching programs of different algorithms that can be played on the camera afterwards or B2B Places for solutions that can be played directly on the camera and only need to be slightly adjusted. The path is going in the perfectly right direction and yet a little time will pass.

How can SMEs still be supported at present?

Basically, a BV system replaces or supports a human control, so when designing and procuring the system, the issue should be addressed in the same way as hiring a skilled worker. It’s about trust, which you give by investing in a system that has to be at least as secure as human testing itself. Therefore, the customer must also be able to understand the system and the technology behind it. SMEs need to be involved in how the optics are selected, how the lighting concept is decided upon and which camera is suitable for the application. The commissioning process must not be hidden in a page-long offer, but must be clearly communicated:

  • We choose this camera system because…
  • We choose this look for this purpose, because…
  • We want to support with this light, due to …
  • We choose for your test this strategy to …
  • The workflow of editing will be as follows, as …

The whole thing takes the abstraction out of Machine Vision – it creates security and understanding. Furthermore, it should be clearly communicated how the project will proceed. Unlike other systems and trades, it is often the case that BV systems are implemented relatively quickly on site in terms of hardware. At the same time, the customer should not have the illusion that everything will immediately work at the desired rate afterwards. Only in the second step can the images decisive for the last 15% of the processing function be collected – which can take between one and four weeks, depending on the plant. After that, the first abstraction of the processing strategy is usually optimized within two to three weeks and a selection rate of at least 98% is achieved in iterative steps. These project steps in particular must be clear in order to increase acceptance of BV systems. When customers are clear about what you’re up against, they have clarity and a fixed timeline for implementation.

And when it comes to the black box Machine Vision itself – it is always advisable to show the individual small steps from the image to the result in the quotation phase:

  • What steps are needed to read an OCR code?
  • What signals do you want to pass on with what results?
  • Which endpoints should be queried during the process and to which should information be passed?
  • How should the documentation of the results look like – Should images also be saved?

So it all seems almost a bit clumsy here at the end, considering the direction you can go in. BV systems are not a mystery, and most importantly, they are often not as expensive as they seem.

The costs are nevertheless an important decision factor

When we did a market research among 100 users and manufacturers in the context of a BV system for CNC milling, the question of how much a camera system would be worth to them often came up with the amount of 5000€-10000€, or from 10,000€ to 15,000€. Maybe it is a bit out of the air, but actually we notice that single BV systems are worth just SMEs up to 15.000€, after that the step to order itself is still too big.

Is this amount possible?

Basically, it always depends on how the system is set up and what you want to check. A distinction must always be made between hardware costs, integration costs and additional development costs:

For hardware, the price per camera system (camera, light and optics) is usually around 6000€ if you want to have Machine Vision within a clock rate of more than one second. If you have a faster clock rate, the price increases significantly because the performance of all adjusting screws, such as camera, computing performance and light, have to be increased equally. It is important to pay attention to which expansion stages of cameras you get. Often, cameras are sold with a full range of functions that are never needed for the application. Is this the trustworthy point mentioned above? Certainly not.

In addition, there are integration costs which, if well planned, can usually cost a maximum of 3000€ per system, often synergies reduce this price, i.e. if more systems have to be set up at the same time. So, in the end, there is still 7000€ left to develop and implement a strategy, if you take the maximum rate of 15.000€.

So “YES”, this can work – Unfortunately only with the exception that it is always dependent on the requirement and must be re-evaluated each time.

What use is it to me as a medium-sized company?

The beauty of integrating camera systems is that you have no impact on the production system during integration. It may be that installation is carried out on non-production days, but it is precisely the initial phase, when the vision system is put into operation, that does not cause any delays in everyday production. The project is running in parallel, as it were, until the release. When the desired target is reached, the system is switched to productive.

Furthermore, a BV system can replace a specialist in a specific task – but it can also support a specialist in specific tasks. The goal is to get employees more involved in the important tasks of the company. Or to let them react quickly and purposefully when the system reports a problem part. At the same time, companies have the opportunity to prepare for macro issues such as the shortage of skilled workers, but also to reduce the prevailing cost pressure, since, for example, the systems can work not only the “40h” during the week, but also the 48h on weekends.

From this point of view, it is an important and correct step to think about how certain tasks should be handed over or supported in your companies in the future.

Other contributions:


Optimize processes: Transformation through modern image processing


10:00 a.m.