EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

BSDR: A Data-Efficient Deep Learning-Based Hyperspectral Band Selection Algorithm Using Discrete Relaxation

Mohammad Rahman iD, Shyh Wei Teng iD, Manzur Murshed iD, Manoranjan Paul iD, David Brennan

Sensors · 04.12.2024 · 24 (23) · 7771

DOI: 10.3390/s24237771 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Sensors, online erschienen am 04.12.2024. Verfasst von Mohammad Rahman und 4 weiteren Autor:innen. Beteiligte Einrichtung: Institute of Innovation, Science and Sustainability, Federation University Australia, University Drive, Mt Helen, VIC 3350, Australia. Die Arbeit wurde bisher 9-mal zitiert (Crossref) und nennt 45 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Soil Cooperative Research Centre, Cooperative Research Centre for High Performance Soils.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Hyperspectral band selection algorithms are crucial for processing high-dimensional data, which enables dimensionality reduction, improves data analysis, and enhances computational efficiency. Among these, attention-based algorithms have gained prominence by ranking bands based on their discriminative capability. However, they require a large number of model parameters, which increases the need for extensive training data. To address this challenge, we propose Band Selection through Discrete Relaxation (BSDR), a novel deep learning-based algorithm. BSDR reduces the number of learnable parameters by focusing solely on the target bands, which are typically far fewer than the original bands, thus resulting in a data-efficient configuration that minimizes training data requirements and reduces training time. The algorithm employs discrete relaxation, transforming the discrete problem of band selection into a continuous optimization task, which enables gradient-based search across the spectral dimension. Through extensive evaluations on three benchmark datasets with varying spectral dimensions and characteristics, BSDR demonstrates superior performance for both regression and classification tasks, achieving up to 25% and 34.6% improvements in overall accuracy, compared to the latest attention-based and traditional algorithms, respectively, while reducing execution time by 96.8% and 97.18%. These findings highlight BSDR’s effectiveness in addressing key challenges in hyperspectral band selection.

Schlagwörter

Stichwörter der Autor:innen

band selectiondata-efficientdiscrete relaxationgradient-based searchhyperspectral

Autor:innen und Institutionen

  • Institute of Innovation, Science and Sustainability, Federation University Australia, University Drive, Mt Helen, VIC 3350, Australia.
  • Cooperative Research Centre for High Performance Soils, Callaghan, NSW 2308, Australia.
  • School of Information Technology, Deakin University, 221 Burwood Hwy, Burwood, VIC 3125, Australia.
  • School of Computing, Mathematics and Engineering, Charles Sturt University, Panorama Ave, Bathurst, NSW 2795, Australia.
  • Wimmera Catchment Management Authority, 24 Darlot St, Horsham, VIC 3400, Australia.

ORCID-Profile: Mohammad Rahman, Shyh Wei Teng, Manzur Murshed, Manoranjan Paul

Förderung

  • Soil Cooperative Research Centre
  • Cooperative Research Centre for High Performance Soils

Zitieren

APA

Rahman, M., Teng, S. W., Murshed, M., Paul, M., & Brennan, D. (2024). BSDR: A Data-Efficient Deep Learning-Based Hyperspectral Band Selection Algorithm Using Discrete Relaxation. Sensors, 24(23), 7771. https://doi.org/10.3390/s24237771

Vancouver

Rahman M, Teng SW, Murshed M, Paul M, Brennan D. BSDR: A Data-Efficient Deep Learning-Based Hyperspectral Band Selection Algorithm Using Discrete Relaxation. Sensors. 2024;24(23):7771. doi:10.3390/s24237771

BibTeX
@article{Rahman2024,
  author = {Rahman, Mohammad and Teng, Shyh Wei and Murshed, Manzur and Paul, Manoranjan and Brennan, David},
  title = {BSDR: A Data-Efficient Deep Learning-Based Hyperspectral Band Selection Algorithm Using Discrete Relaxation},
  journal = {Sensors},
  year = {2024},
  volume = {24},
  number = {23},
  pages = {7771},
  doi = {10.3390/s24237771}
}
Kostenlos starten →