View in EDS

Cognitively inspired deep learning for hyperspectral remote sensing-based land cover classification.

Saved in:
Bibliographic Details
Title: Cognitively inspired deep learning for hyperspectral remote sensing-based land cover classification.
Authors: S.M, Devendra Kumar1 (AUTHOR) smdevendrakumar15@gmail.com, M., Devanathan2 (AUTHOR) devanathan.m@reva.edu.in
Source: Multimedia Tools & Applications. Jul2026, Vol. 85 Issue 7, p1-29. 29p.
Abstract: The spectral signatures consist of fine-grained information that can be obtained using Hyperspectral Imaging (HSI), and hence it is of utmost importance in the area of remote sensing applications. Various Artificial Intelligence (AI) based approaches, viz. Machine Learning (ML) and Deep Learning (DL) have evolved towards performing analysis of critical and complex attributes in HSI. Conventional ML/DL approaches depend on unified processing of spectral and spatial dimensions, thereby generating degraded representations. Proposed study, in line with addressing this open-ended challenge, introduces an analytical framework using DL that is progressively optimized by cognitive capabilities towards classifying landcovers. The proposed study investigated comparing three explicit architectures of deep learning, viz., (i) Simple Deep Learning (SDL) deployed as a 2D Convolutional Neural Network (2D-CNN), (ii) Cognitive Deep Learning (CDL) deployed as a volumetric 3D Convolutional Neural Network (3D-CNN), and (iii) Advanced Cognitive Deep Learning (ADCL) extending the 3D-CNN backbone. SDL processes spectral bands as multi-channel spatial inputs while CDL jointly models spectral–spatial correlations, and ADCL incorporates residual feature propagation and pyramid pooling towards improving multi-scale spatial aggregation. Different from any attention-centric band weighting strategies, the proposed ADCL model mechanises spatial context fusion and hierarchical residual learning for enhancing discriminative representation. Experimented on the Pavia University dataset, ADCL scored 99.34% overall accuracy with 0.9913 Cohen’s kappa score. [ABSTRACT FROM AUTHOR]
Copyright of Multimedia Tools & Applications is the property of Springer Nature and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)
Database: Engineering Source
Description
Abstract:The spectral signatures consist of fine-grained information that can be obtained using Hyperspectral Imaging (HSI), and hence it is of utmost importance in the area of remote sensing applications. Various Artificial Intelligence (AI) based approaches, viz. Machine Learning (ML) and Deep Learning (DL) have evolved towards performing analysis of critical and complex attributes in HSI. Conventional ML/DL approaches depend on unified processing of spectral and spatial dimensions, thereby generating degraded representations. Proposed study, in line with addressing this open-ended challenge, introduces an analytical framework using DL that is progressively optimized by cognitive capabilities towards classifying landcovers. The proposed study investigated comparing three explicit architectures of deep learning, viz., (i) Simple Deep Learning (SDL) deployed as a 2D Convolutional Neural Network (2D-CNN), (ii) Cognitive Deep Learning (CDL) deployed as a volumetric 3D Convolutional Neural Network (3D-CNN), and (iii) Advanced Cognitive Deep Learning (ADCL) extending the 3D-CNN backbone. SDL processes spectral bands as multi-channel spatial inputs while CDL jointly models spectral–spatial correlations, and ADCL incorporates residual feature propagation and pyramid pooling towards improving multi-scale spatial aggregation. Different from any attention-centric band weighting strategies, the proposed ADCL model mechanises spatial context fusion and hierarchical residual learning for enhancing discriminative representation. Experimented on the Pavia University dataset, ADCL scored 99.34% overall accuracy with 0.9913 Cohen’s kappa score. [ABSTRACT FROM AUTHOR]
ISSN:13807501
DOI:10.1007/s11042-026-21741-5