Initial Situation, Problem Definition or Motivation

European industry is undergoing a rapid transition toward greener and smarter manufacturing, while sectors such as photovoltaics and microelectronics are expanding significantly. Even small yield fluctuations of 1–3% can already result in losses amounting to millions of euros. At the same time, existing measurement methods are reaching their limits: point spectroscopy measures only isolated locations, full-field hyperspectral techniques are too slow, and conventional machine vision remains almost “chemically blind” despite very high frame rates. As a result, critical defect modes such as contamination, layer-thickness variations, microcracks or porosity often go undetected. This highlights a fundamental conflict between spatial and spectral resolution, speed, and chemical sensitivity—one that continues to hinder reliable material diagnostics at industrial scale.

Goals and Innovation Content

This cooperative R&D project directly responds to the call topic “Laser-based Quality Control. ” Its aim is to establish a scientifically grounded industrial research line that investigates a novel measurement principle for spectral inline inspection. At its core is the combination of Coded Aperture Snapshot Spectral Imaging (CASSI), a technique capable of capturing a complete hyperspectral data cube in a single coded exposure, with the Fisher Information (FI) concept. FI quantifies how much usable information a measurement provides about a given material property; put simply, how well even the smallest differences can be distinguished. The higher the FI, the more precisely minimal changes can be detected and the more powerful a measurement system can become. By using FI-optimised optical codes and targeted sampling strategies, the project aims to achieve more than a tenfold reduction in data volume (around 10 percent of conventional data) while maintaining comparable spatial and spectral resolution, and simultaneously enabling the millisecond-level frame rates required for laser-based processes. The central scientific premise is that the resolution of a measurement system is not a fixed property of the sensor, but can be actively engineered by optimising information flow within both the optical setup and the downstream reconstruction. This establishes a new, physically grounded pathway toward ultrafast and chemically sensitive spectroscopy.Desired results or findings The project will experimentally investigate whether CASSI coding, FI-guided sampling and physics-based AI can together form a unified, information-driven measurement principle that surpasses the performance of classical, heuristically developed hyperspectral approaches. The guiding hypothesis is that spectral resolution, robustness and chemical sensitivity are not merely given, but can be actively designed. Under controlled laboratory conditions, all coded measurement data will be reconstructed entirely on edge hardware using lightweight, physics-based AI, while only compact quality indicators are output, in line with data minimisation principles and the requirements of the AI Act. Scientifically ambitious yet methodologically accessible, the project provides fundamental insights into information- driven spectroscopy and paves the way for future generations of ultrafast hyperspectral measurement systems in European industry.