Synthesis of Simplified Spectral Imaging Systems: Methodology for Selecting Key Bands Using Weight Analysis and Effectiveness Thresholds
IEEE Access · 2026 · 14 · 114347-114355
DOI: 10.1109/access.2026.3714764 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Access, 2026. Verfasst von Jedrzej Kowalewski und 2 weiteren Autor:innen. Die Arbeit nennt 30 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Ministry of Education and Science.
Abstract (Originalsprache)
In this paper, we have demonstrated that task-controlled spectral band selection enables the synthesis of simple, low-cost multispectral imaging (MSI) systems that maintain high classification performance despite a radical reduction in the number of channels. Using four canonical hyperspectral datasets (Indian Pines, Salinas, Salinas-A, Pavia Centre), selected for their diversity of scenes and spectral-spatial parameters, we showed that the sets of bands determined by selected data processing algorithms like Principle Component Analysis and Linear Discriminant Analysis, and Random Forest are largely complementary (low IoU/Jaccard, $\le \sim 30$ %), and $\ge 90$ -95% effectiveness can be achieved with as few as 3-5 bands. Most importantly, our research showed that correct classification is feasible with as few as ~ one-third of the ground-truth classes, challenging the common heuristic that the band count should match or exceed the number of classes. The developed procedure (band ranking $\to $ “accuracy vs. number of bands” curves $\to $ mapping to filter parameters) creates a practical design path for dedicated MSI instruments based on simpler data acquisition systems, more favorable radiometry, and lower implementation costs due to the reduction of the number of channels, which results in less required data.
Förderung
- Ministry of Education and Science
Zitieren
APA
Kowalewski, J., Jasinska, L., & Domaradzki, J. (2026). Synthesis of Simplified Spectral Imaging Systems: Methodology for Selecting Key Bands Using Weight Analysis and Effectiveness Thresholds. IEEE Access, 14, 114347-114355. https://doi.org/10.1109/access.2026.3714764
Vancouver
Kowalewski J, Jasinska L, Domaradzki J. Synthesis of Simplified Spectral Imaging Systems: Methodology for Selecting Key Bands Using Weight Analysis and Effectiveness Thresholds. IEEE Access. 2026;14:114347-114355. doi:10.1109/access.2026.3714764
BibTeX
@article{Kowalewski2026,
author = {Kowalewski, Jedrzej and Jasinska, Laura and Domaradzki, Jaroslaw},
title = {Synthesis of Simplified Spectral Imaging Systems: Methodology for Selecting Key Bands Using Weight Analysis and Effectiveness Thresholds},
journal = {IEEE Access},
year = {2026},
volume = {14},
pages = {114347--114355},
doi = {10.1109/access.2026.3714764}
}
EUVIMED