Multi-stage image processing approach for 3D-CT based in- or at-line inspection of battery cells.
e-Journal of Nondestructive Testing · 2026-07 · 31 (7)
DOI: 10.58286/33089 ↗Kurz gefasst
Fachartikel in der Zeitschrift e-Journal of Nondestructive Testing, 2026-07. Verfasst von S. Gondrom-Linke und 2 weiteren Autor:innen.
Abstract (Originalsprache)
In this contribution, we will introduce our multi-stage image processing approach which combines analytical approaches with multiple, robust CNN based solutions, such as classification networks and anomaly detection networks, and is used e.g. for inline 3D-CT inspection of battery cells. We will discuss the advantages and disadvantages of the different methods and highlight a special kind of in-house developed anomaly detection. We will pay particular attention to networks that require only a small amount of data for training and are also insensitive to suboptimal segmentation of defects in the training data. Only this ensures real applicability in an industrial production environment. We will show that this combination of different highly sophisticated approaches to solve complex inspection problems, such as detection of very small inclusions in battery cells, together with intelligent pre- and postprocessing of the data, helps us to overcome previous limits and restrictions in analysis and the handling of image processing.
Zitieren
APA
Gondrom-Linke, S., Ramos, T., & Schmid, S. (2026). Multi-stage image processing approach for 3D-CT based in- or at-line inspection of battery cells. e-Journal of Nondestructive Testing, 31(7). https://doi.org/10.58286/33089
Vancouver
Gondrom-Linke S, Ramos T, Schmid S. Multi-stage image processing approach for 3D-CT based in- or at-line inspection of battery cells. e-Journal of Nondestructive Testing. 2026;31(7). doi:10.58286/33089
BibTeX
@article{GondromLinke2026,
author = {Gondrom-Linke, S. and Ramos, Tiago and Schmid, S.},
title = {Multi-stage image processing approach for 3D-CT based in- or at-line inspection of battery cells},
journal = {e-Journal of Nondestructive Testing},
year = {2026},
volume = {31},
number = {7},
doi = {10.58286/33089}
}
EUVIMED