EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Augmenting Smoke Detection Systems via Radar and Deep Learning for Continuous Hazard and Condition Monitoring: A Multispectral Analysis

Robin Schmitz, Francesca Schenkel, Tobias Körner, Thorsten Schultze, Christoph Baer, Jan C. Balzer, Jan Barowski, Ilona Rolfes, Christian Schulz

IEEE Transactions on Radar Systems · 2026 · 4 · 787-798

DOI: 10.1109/trs.2026.3682428 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Transactions on Radar Systems, 2026. Verfasst von Robin Schmitz und 8 weiteren Autor:innen. Die Arbeit nennt 48 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Deutsche Forschungsgemeinschaft through the Project Mobile Material-Charakterisierung und -Ortung durch Elektromagnetische Abtastung (MARIE) and BULK-REACTION.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Traditional smoke detection systems share inherent limitations. Inter alia, these systems are fundamentally limited by their purely reactive and nonpredictive operation. The objective of this contribution is to extend the advantageous probing capabilities of radar systems toward the augmentation of traditional smoke detectors for continuous condition and hazard monitoring. The particular emphasis is placed here on early stage hazards and critical precursor conditions for preventive fire safety. To this end, static-target radar echoes are classified using deep learning (DL) methods by extracting characteristic patterns caused by subtle changes in the surrounding medium (e.g., water vapor). Measurements were conducted employing ultrawideband frequency-modulated continuous-wave (FMCW) radar systems operating from 80 to 240 GHz, in a controlled indoor proof-of-concept setup, enabling a systematic multispectral analysis. Several scenarios representative of domestic settings were considered, including the frying of a burger patty (BP) (UL 268 standard). The proposed concept achieves classification accuracies of more than 90%, even under strong restrictions. The developed models remain lightweight and fast, well-suited for embedded implementations.

Förderung

  • Deutsche Forschungsgemeinschaft through the Project Mobile Material-Charakterisierung und -Ortung durch Elektromagnetische Abtastung (MARIE) and BULK-REACTION

Zitieren

APA

Schmitz, R., Schenkel, F., Körner, T., Schultze, T., Baer, C., Balzer, J. C., Barowski, J., Rolfes, I., & Schulz, C. (2026). Augmenting Smoke Detection Systems via Radar and Deep Learning for Continuous Hazard and Condition Monitoring: A Multispectral Analysis. IEEE Transactions on Radar Systems, 4, 787-798. https://doi.org/10.1109/trs.2026.3682428

Vancouver

Schmitz R, Schenkel F, Körner T, Schultze T, Baer C, Balzer JC, et al. Augmenting Smoke Detection Systems via Radar and Deep Learning for Continuous Hazard and Condition Monitoring: A Multispectral Analysis. IEEE Transactions on Radar Systems. 2026;4:787-798. doi:10.1109/trs.2026.3682428

BibTeX
@article{Schmitz2026,
  author = {Schmitz, Robin and Schenkel, Francesca and Körner, Tobias and Schultze, Thorsten and Baer, Christoph and Balzer, Jan C. and Barowski, Jan and Rolfes, Ilona and Schulz, Christian},
  title = {Augmenting Smoke Detection Systems via Radar and Deep Learning for Continuous Hazard and Condition Monitoring: A Multispectral Analysis},
  journal = {IEEE Transactions on Radar Systems},
  year = {2026},
  volume = {4},
  pages = {787--798},
  doi = {10.1109/trs.2026.3682428}
}
Kostenlos starten →