P‐5.1: A Class‐Balanced Spectral‐Spatial Residual Network for Fine‐Grained Hyperspectral Vegetation Classification
SID Symposium Digest of Technical Papers · 2026-08 · 57 (S1) · 1304-1309
DOI: 10.1002/sdtp.70003_347 ↗Kurz gefasst
Fachartikel in der Zeitschrift SID Symposium Digest of Technical Papers, online erschienen am 11.09.2026. Verfasst von Ruiyang Fei. Die Arbeit nennt 18 Literaturangaben.
Abstract (Originalsprache)
Hyperspectral remote sensing imagery (HSI), characterized by contiguous narrow spectral bands and rich spatial information, offers strong potential for fine‐grained vegetation classification. However, in real‐world scenarios, highly similar spectral signatures among tree species, severe class imbalance, and complex canopy structures make it difficult for traditional hand‐crafted feature‐based methods such as random forest and support vector machine to effectively exploit spectral‐spatial information, resulting in frequent confusion among spectrally similar and minority classes. To address these challenges, we propose a Class‐Balanced Spectral‐Spatial Residual Network (CB‐SS‐ResNet) that adopts a dual‐branch architecture to jointly model spectral details and spatial textures, incorporates vegetation indices and texture features as auxiliary inputs, and employs class‐balanced sampling with hard‐sample‐focused optimization to enhance minority‐class learning. Experiments on the Xiongan Matiwan airborne hyperspectral dataset demonstrate that the proposed method improves discrimination of spectrally similar and minority classes while maintaining high overall accuracy and Kappa coefficient, effectively reducing confusion among species such as Robinia, pear, and poplar.
Zitieren
APA
Fei, R. (2026). P‐5.1: A Class‐Balanced Spectral‐Spatial Residual Network for Fine‐Grained Hyperspectral Vegetation Classification. SID Symposium Digest of Technical Papers, 57(S1), 1304-1309. https://doi.org/10.1002/sdtp.70003_347
Vancouver
Fei R. P‐5.1: A Class‐Balanced Spectral‐Spatial Residual Network for Fine‐Grained Hyperspectral Vegetation Classification. SID Symposium Digest of Technical Papers. 2026;57(S1):1304-1309. doi:10.1002/sdtp.70003_347
BibTeX
@article{Fei2026,
author = {Fei, Ruiyang},
title = {P‐5.1: A Class‐Balanced Spectral‐Spatial Residual Network for Fine‐Grained Hyperspectral Vegetation Classification},
journal = {SID Symposium Digest of Technical Papers},
year = {2026},
volume = {57},
number = {S1},
pages = {1304--1309},
doi = {10.1002/sdtp.70003_347}
}
EUVIMED