Five Machine Vision Technologies Worth Watching

Five innovations have been shortlisted for the VISION Award 2026, to be presented on 7 October at VISION 2026 in Stuttgart. Award shortlists are usually marketing, but this one is a reasonable map of where industrial imaging is going, and two of the five address problems that limit vision projects today rather than adding resolution to ones that already work.

Lidwave's Odem is a 4D coherent vision sensor that measures range, reflectivity and instantaneous velocity at every pixel, using Doppler information rather than comparing consecutive frames — with the optical engine integrated onto a single silicon chip. Per-pixel velocity from Doppler is a categorically different measurement from velocity inferred by tracking a feature between frames: it does not need the feature, it does not accumulate error over the frame interval, and it works on the first frame. For anything involving moving objects and safety-relevant decisions, that changes the latency budget.

Medabsy's Virtualising Machine Vision attacks the other end of the problem. It imports 3D component models, configures imaging hardware and simulates inspection results before any physical system is built, and generates photorealistic synthetic datasets with pixel-level defect annotations for machine-learning training. Two of the most expensive failure modes in vision projects are discovering after installation that the chosen optics and lighting cannot see the defect, and being unable to train a classifier because the defect is rare enough that the plant has forty examples. Simulating the first and synthesising the second is aimed squarely at both.

The other three are further upstream. AIT Austrian Institute of Technology entered PHOTODEX, a mobile inspection system combining high-speed imaging with colour-coded strobing to capture photometric data and register detected defects against CAD models, linking surface inspection to digital twins. photonicSENS entered Industrial 3D Light Field Vision, a plenoptic technology capturing 2D images and depth simultaneously from a single sensor and a single exposure with no projected pattern — which removes the projector, the alignment and the ambient-light fight that structured light brings with it. Singular Photonics entered Litavis, a software-configurable SPAD image sensor combining single-photon detection with on-chip digital photon processing, aimed at machine vision, robotics, LiDAR and industrial automation.

The common thread is worth naming: four of the five reduce what has to be true about the scene before a measurement works — no projected pattern, no frame-to-frame tracking, no physical prototype, no photon budget. Vision systems fail in production far more often because of the environment than because of the algorithm, and that is where the shortlist is pointing.