, ,

VISION 2026: A Machine Vision Industry Ready to See the Bigger Picture

Visitors networking among machine vision exhibitor stands at VISION 2026 in Stuttgart.

From award winning imaging technology and 200GigE cameras to more open systems and the growing influence of AI, VISION 2026 offered plenty to consider. MV Pro Media reflects on the demonstrations, discussions and people that made the show.

There is something about Vision that makes it difficult to leave with quite the same view of the machine vision industry that you arrived with.

We spend much of the year reading announcements, studying product specifications and speaking to companies about their latest developments. Yet there is no substitute for walking the halls in Stuttgart, seeing technologies in action and hearing directly from the people developing and using them.

This year, those conversations felt particularly significant. Artificial intelligence was everywhere, but so were discussions about optics, lighting, sensors, networking and the practical work of making a vision system perform reliably. For every demonstration that encouraged us to imagine what might be possible in the future, there was another focused on making an existing application easier to build, deploy or maintain.

For MV Pro Media, VISION 2026 was a special show. We arrived with our VISION special edition, a booth, a full schedule of interviews and the opportunity to host a panel discussion on one of the industry’s most talked about subjects. Across three days, we met established leaders, ambitious start ups, engineers, researchers and people discovering machine vision for the first time.

What stayed with us was the sense of an industry thinking more deeply about the complete journey, from capturing light and moving data to making a decision and, increasingly, taking action.

The MVPro Media team at VISION 2026 in Stuttgart.

A record show with an international outlook

The numbers tell part of the story. Held at Messe Stuttgart from 6 to 8 October, Vision 2026 attracted a record 494 exhibitors and more than 9,000 trade visitors, representing a 14 per cent increase in attendance compared with 2024. Visitors came from 59 countries, while around two thirds of exhibitors were based outside Germany.

That international character was clear on the exhibition floor. Conversations moved between manufacturing challenges in Europe, automation opportunities in North America and developments from imaging companies across Asia. Individual markets face different pressures, but many of the underlying questions are shared: how do we inspect more effectively, deploy systems more easily and make better use of the information that vision provides?

There was also energy around the show’s newer formats. The start up programme gave emerging companies a prominent platform, guided tours helped visitors explore subjects including AI, and the first Vision Makeathon brought students and apprentices together to tackle practical machine vision challenges. Alongside the Industrial Vision Days presentations, these activities gave visitors reasons to look beyond individual stands and engage with the ideas shaping the sector.

For us, that combination is what makes Vision so valuable. It is a place to discover products, but also somewhere to test assumptions, encounter unfamiliar perspectives and understand how the industry is responding to changing customer needs.

From individual components to complete applications

One of the strongest themes to emerge from Stuttgart was a growing focus on how the pieces of a vision system fit together.

Machine vision has always depended on the relationship between camera, sensor, lens, lighting, software and processing. Yet buyers have often been asked to navigate those elements separately, bringing together products from different suppliers and taking responsibility for making them work as a dependable whole.

At Vision, we saw more companies presenting their technologies in the context of the application. The conversation was increasingly about the image a customer needs, the conditions in which it must be captured and the outcome the system is expected to deliver.

Allied Vision offered a clear example. Following the integration of several established 2D machine vision brands under the Allied Vision name, the company presented a broader ecosystem spanning cameras, lenses, frame grabbers, software libraries and interfaces. The aim is to make it easier for customers to bring compatible components together and obtain support across more of the imaging chain.

Examples involving Basler, CCS and Teledyne DALSA also highlighted the importance of treating illumination and image acquisition as connected parts of the application. A camera’s performance cannot be judged in isolation from the light falling on the object, the lens forming the image or the consistency of the environment in which the system operates.

LUCID Vision Labs showed another approach with its Triton All in One camera, bringing a lens, programmable illumination and a camera into an IP67 housing. Its liquid lens can adjust focus as working distances change, while processing remains external, allowing customers to select their own computing hardware and software. At the time of our interview, production was planned for early 2027.

None of these approaches removes the need for application expertise. They do, however, suggest that suppliers are taking greater responsibility for making their technologies easier to combine and put to work. That could mean fewer surprises during commissioning and a faster route to a system that delivers the required result.

The award that brought a new perspective to 3D

Among the show’s biggest moments was the presentation of the Vision Award to photonicSENS for its apiCAM light field imaging technology.

The company’s approach uses a microlens array to capture information from multiple angular perspectives within a single exposure. Algorithms then use that information to produce a 2D image and a corresponding depth map from one sensor. The attraction is the possibility of obtaining depth information without a separate projector or a mechanical scan, although performance must ultimately be judged against the conditions and requirements of each application.

What makes the award interesting is not simply that another approach to 3D imaging has entered the conversation. It is the way photonicSENS has taken a long established optical principle and worked towards making it practical for industrial use.

Three dimensional vision is no longer a category with one obvious answer. Stereo vision, structured light, laser triangulation, time of flight and other approaches each bring strengths and limitations. The most useful development is often the one that gives an engineer another viable option when faced with restricted space, moving objects, challenging geometry or difficult surfaces.

The other shortlisted entries reflected that breadth. AIT Austrian Institute of Technology, Lidwave, Medabsy and Singular Photonics joined photonicSENS on the shortlist, with work spanning different areas of imaging and system development.

Start ups brought fresh ideas to the floor

The start up programme showed how much innovation is taking place across the imaging chain.

Phaseform, eyeo imaging and TiHive were each recognised as Vision Start up of the Day, before eyeo went on to win the overall VisionStart up 2026 award. For eyeo, it was a particularly memorable first appearance at the show: the company said it left Stuttgart with both a daily award and the overall title.

Their technologies address different challenges. Phaseform is working with adaptive optics, which can manipulate a wavefront to correct optical aberrations. TiHive is applying terahertz technology to inspection, exploring ways to investigate materials and structures that may not be readily visible with conventional imaging. eyeo is developing colour sensing technology intended to improve how light is used by image sensors.

We found it encouraging to see emerging companies given a prominent place alongside the industry’s established names. Some of the most enjoyable conversations at VISION began by stopping at a smaller stand and asking a simple question: what problem are you trying to solve?

Faster cameras and smarter data paths

If there was one set of specifications guaranteed to attract attention, it came from Emergent Vision Technologies.

The company used Vision to introduce its 200GigE HYPERION area scan and STRATOS line scan camera platforms. The first HYPERION model, the AH-21000-G, is specified with a 21 megapixel Gpixel sensor and a frame rate of up to 1,000 frames per second. The planned STRATOS LST-16KG5 combines 16K resolution with time delay integration and a line rate of up to 1 MHz.

Those figures become more meaningful when considered alongside the applications behind them. Semiconductor inspection, electronics manufacturing, high speed production and scientific imaging can place extraordinary demands on the amount of visual information a system must acquire and process.

A faster connection is only one part of the solution. Data must also move efficiently into processing hardware, and the wider system needs to manage compute, memory, storage and heat. Emergent’s support for GigE Vision 3.0, RDMA and direct transfers to GPU memory points towards the importance of the complete data path, rather than just the speed written on the camera.

We saw the same principle from another perspective at LUCID Vision Labs. One demonstration connected 20 10GigE cameras through a network with two 100GigE uplinks, using RDMA to transfer images into GPU memory while reducing unnecessary data handling by the CPU. The cameras also needed to capture images at coordinated times, with a shared clock and LUCID’s eTrigger functionality supporting synchronised acquisition across the array.

At another station, LUCID used its 105 megapixel Atlas25 camera to show fine detail on a semiconductor wafer. The camera uses a 25GigE interface and the Sony IMX937 sensor, with the demonstration illustrating both the detail available and the optical and data transfer considerations that accompany such a large image. The 20 camera array and the wafer demonstration were separate examples, but together they made the same point: resolution and interface speed only become valuable when the rest of the system can use the resulting data.

Emergent also presented its AEOLUS camera with onboard H.265 compression, offering a different response to the data challenge. Rather than transmitting every image in its original form, compression can reduce bandwidth and storage requirements. Its suitability depends on the application and what image information must be preserved, but it shows that higher performance does not always require simply building a bigger pipe.

Looking beyond conventional imaging

Another technology attracting attention was event based vision. Rather than capturing complete images at fixed intervals, an event sensor reports changes in brightness at individual pixels as they occur.

In a scene where little is changing, the sensor does not need to generate the same volume of repeated image information as a conventional frame based camera. In fast moving applications, it can also capture changes with very low latency. The resulting data differs from a traditional image, however, so applications need software and processing designed to make use of it.

At Vision, IDS and Prophesee demonstrated event based imaging and its potential for applications where speed, timing and data efficiency matter. It is not a replacement for every conventional camera; it is another way of observing a scene when a particular problem calls for it.

Moving beyond conventional imaging was not limited to the way information is captured over time. During our visit to Anteryon, the discussion around its SiL technology and hyperspectral imaging highlighted another route towards extracting information that a conventional RGB image may not reveal. By examining a scene across more wavelengths, spectral imaging can help distinguish materials or characteristics that appear similar to the human eye.

Fraunhofer IOF provided a different example of vision working in a demanding environment. Its goCRASH3D system combines two cameras with pattern illumination for high speed 3D capture inside a crash test vehicle. The institute describes a system capable of recording up to 1,200 3D images per second while withstanding accelerations of up to 60g.

It is an example of a system having to do more than produce an impressive image. It must also survive the physical conditions in which that information is needed.

Making vision easier to develop

Some of the most consequential demonstrations in Stuttgart focused on the people developing applications.

MVTec’s AI Vision Solver was one example. Presented as a prototype, it combines the company’s HALCON software and HDevelopEVO development environment with AI agents. Users can describe a task using natural language and images, with the agents selecting relevant operators and generating an application for tasks such as code reading, object matching or optical character recognition.

The important word is prototype. It offers a view of how AI could assist development, rather than proof that the work of a vision engineer can be reduced to a prompt.

MVTec emphasises that the final assessment remains with the user. A generated workflow still needs to be evaluated against real images, production variability, acceptable error rates and the consequences of a wrong decision. The opportunity is to reduce some of the routine development work, giving engineers more time to define the problem and validate the result.

That pressure to make vision easier is also commercial. In our conversation with Cognex President and CEO Matt Moschner, he described a market in which barriers to adopting machine vision have fallen while expectations around responsiveness, deployment speed and return on investment continue to rise. For suppliers, making a technology possible is increasingly only the beginning: customers also want to understand how quickly it can be installed and deliver value.

Simulation offers another way to address that challenge. Basler and MVTec highlighted a workflow in which virtual camera streams can be used with HDevelop to develop and test an application before physical hardware is available.

The idea connected with our conversation with Medabsy’s Max Grau and Petra Gospodnetić, who demonstrated software designed to help users test a proposed machine vision system before ordering its components. Being able to explore a setup earlier could help teams identify problems before they become expensive changes on the factory floor.

Real world testing remains essential. Lighting, surfaces, tolerances and environmental conditions can reveal details that a model has missed. But moving some of the learning earlier in a project could make machine vision a less daunting purchase and a more predictable investment.

A more open machine vision ecosystem

Ease of development was only part of the accessibility conversation in Stuttgart. Alongside more complete solutions, we heard calls to make cameras, software and information easier to use across different suppliers’ systems.

Kurokesu’s emphasis on open-source drivers gave that discussion a practical focus: helping developers integrate cameras using familiar tools and retain the freedom to adapt their applications. Our conversation with tekom about iiRDS extended the theme to technical information, showing how recognising a spare part could help users find the documentation needed to work with it.

What connected these conversations was a broader view of usability. Getting a system running is one milestone; maintaining it, adding capabilities and connecting it to other equipment are part of the same customer journey. As vision takes on a greater role in automation and robotics, those longer-term considerations deserve attention alongside deployment speed.

For us, this was another sign of the industry thinking beyond individual products towards how they will be used throughout their working lives. It also raised questions about the relationship between integration and customer choice, which we explore in Does Machine Vision Need to Be More Open?.

From seeing to acting

AI was impossible to avoid at VISION, but the most useful discussions went beyond asking whether a product included it.

One of the highlights for MV Pro was chairing the VDMA panel, “From Seeing to Acting: Is Agentic AI Reinventing Machine Vision?” The discussion brought together perspectives from AUMOVIO, Intrinsic, IDS and Siemens to consider how vision, AI and automation are beginning to work together in new ways.

The idea of agentic AI raises an interesting possibility for machine vision. A system might not simply identify an object or flag a defect; it could use visual information as part of a wider process that selects tools, considers context and supports a subsequent action.

Robotics gives that discussion a practical setting. In our interview, Matt Moschner discussed Cognex’s proposed acquisition of RealSense, which was pending at the time, and the different demands placed on robotic perception. Mobile and increasingly autonomous robots may need to operate across wider fields of view, longer working distances and changing lighting conditions, rather than looking at a predictable inspection point on a production line.

As the ambition grows, so do the questions. How does a system recognise uncertainty? How can its decisions be tested and explained? Who is responsible when an action affects a production process? And how do we ensure that a promising demonstration remains reliable after weeks or months of operation?

We came away from the panel feeling that agentic AI is a subject the industry needs to engage with seriously, while retaining the engineering discipline that has made machine vision valuable. The opportunity is substantial, but trust will be earned through performance, transparency and evidence.

Optics still matter

With so much discussion around AI and data, it would be easy to assume that the future of machine vision belongs entirely to software. A walk around VISION quickly corrected that impression.

Our visit to poLight was a good example. Speaking with Jon Edwards about the company’s TLens technology and MLens products brought us back to a fundamental challenge: getting the right image in focus when the working distance or object position changes.

A tunable lens can provide a way to adjust focus without moving a conventional lens assembly in the same way as a traditional mechanical autofocus system. Where objects arrive at varying heights or multiple focal positions need to be inspected, that flexibility can influence the design of the complete application.

Optics and AI are not competing approaches. Better image formation can make a software task easier, while more capable processing can help extract useful information from challenging images. Lighting, filters and sensor choice remain just as important to whether the system can see what it needs to see.

The people behind the technology

Among the product demonstrations, awards and panel discussions were the moments that make a trade show enjoyable.

At MVTec, visitors could discuss machine vision over a freshly prepared coffee and return for a beer after the show. Even the AI Vision Solver demonstrations had a distinctly local flavour, with pretzel classification among the application examples. It is difficult to imagine a more appropriate way to demonstrate machine learning in Stuttgart.

The atmosphere around the halls showed why these events continue to matter. A conversation that begins with a camera specification can quickly turn into a discussion about a difficult inspection problem, a new partnership or an idea someone has been waiting for the right person to help develop.

For the MV Pro team, those conversations were at the heart of our time at VISION. From interviews at company booths to discussions at our own stand, it was a pleasure to hear directly from the people building the products, asking the questions and shaping the industry’s future.

What happens next?

Machine vision is moving in several directions at once. Cameras are becoming faster, while other systems aim to capture or transmit data more intelligently. AI is changing how applications can be developed, while experienced engineers remain essential to testing and validating them. Suppliers are bringing components together, even as open drivers, shared standards and new sensing technologies expand the choices available.

For us, the central challenge is making that progress useful. Customers do not buy megapixels, neural networks or interfaces for their own sake. They buy the ability to inspect a product, guide a robot, understand a process or make a decision with greater confidence.

The demonstrations and discussions we encountered at VISION showed how many ways there are to reach those outcomes. A better data path can make high resolution inspection practical. Simulation can help teams assess a design before buying hardware. An open driver can make integration easier. Better optics and lighting can provide an image that the software has a stronger chance of interpreting.

VISION 2026 gave us plenty of reasons to be optimistic about what is possible. It also reminded us that the future of machine vision will be shaped by conversations between the people who create the technology and the people who need it to work.

To everyone who stopped by the MV Pro booth, joined us for an interview, took part in the panel or simply shared a conversation with us in Stuttgart: thank you. It was a brilliant few days, and we are already looking forward to continuing those conversations.

Most Read

Related Articles

Sign up to the MVPro Newsletter

Subscribe to the MVPro Newsletter for the latest industry news and insight.

Name
Consent

Trending Articles

Latest Issue of MVPro Magazine

MVPro Media
Privacy Overview

This website uses cookies so that we can provide you with the best user experience possible. Cookie information is stored in your browser and performs functions such as recognising you when you return to our website and helping our team to understand which sections of the website you find most interesting and useful.