Automated image processing

Individualized evaluation of optical images for distinctive properties

Image processing is an essential part of optical metrology, because any measuring system based on camera technology not only needs to capture images, but also to evaluate them.

The Fraunhofer IPT therefore researches and develops new image processing evaluation methods for metrological tasks. These include classical approaches such as object recognition as well as dimensional measurement and defect classification, but also current approaches based on deep learning methods.

In automated cell cultivation plants, it is necessary to take large microscope images of stem cells in order to calculate key cell culture data such as confluence, location and size of individual colonies. The result of the classification, which is obtained using artificial intelligence approaches, is not only more accurate and robust against external influences, but is also available much faster, thus enabling efficient process control.

For quality control of technical components, image processing algorithms can detect and evaluate errors such as shape and position deviations. The Fraunhofer IPT develops individual algorithms for such automated manufacturing processes and integrates them into existing imaging and processing systems.

Our services

  • Conception and development of test software based on deep learning algorithms
  • Development of algorithms for quality assurance
  • Development of image processing for process control
  • Integration in plants and manufacturing processes


Cell biology

Algorithms based on deep learning methods allow the classification of cell cultures in automated production with regard to confluence, location and size as well as the evaluation of the growth process for decisions on further exploitation.


In addition to standard methods such as dimensional measurement and detection of surface defects, defect locations as well as position, shape or short circuits, we also develop individual evaluation methods based on artificial intelligence and integrate these into production facilities.


Tomographic measurements such as optical coherence tomography enable image evaluation for process control in manufacturing processes from plastic welding to biological laboratory processes.



In the StemCellFactory, a neuronal network takes over the rapid classification of stem cells.

Ceramic test bed

Image processing is used to detect, measure and evaluate defects in transparent components.


Very large, homogeneously illuminated images are composed of many individual shots and made possible by real-time stitching in combination with shading corrections and histogram spread.



(Piotrowksi et al., 2021): Deep-learning-based multi-class segmentation for automated, non-invasive routine assessment of human pluripotent stem cell culture status, Fraunhofer IPT, 2021.