Hyperspectral Image Classification using Archimedes Optimized 3D CNN layers

Title

Hyperspectral Image Classification using Archimedes Optimized 3D CNN layers

Creator

Athul Ghosh, P.; Chidambaram, S.

Description

Background: Hyperspectral Image (HSI) classification involves analyzing images captured across numerous spectral bands to identify and categorize materials or objects. By exploiting spectral information, HSI enables precise detection of materials based on their unique spectral signatures. This extensive spectral information allows for accurate identification of materials and land cover types, making HSI classification essential for various applications. This research developed an optimization-based Deep Learning (DL) framework for effective classification of HSI. The methodology begins with a preprocessing step using Minimum Noise Fraction (MNF), a technique that leverages the noise covariance structure of the dataset to reduce dimensionality while preserving the most informative spectral features. To enhance classification performance, a novel model is proposed, known as Archimedes Optimization Algorithm-3D Convolutional Neural Network (AOA based 3D CNN). Archimedes Optimization Algorithm (AOA) is applied to fine-tune learning rate of 3D Convolutional Neural Network (3D CNN), thereby optimizing training convergence and enhancing accuracy. Result: The proposed AOA based 3D CNN approach exhibits outstanding performance, achieving 95.6% accuracy, 95.9% TPR, 94.1% TNR, 94.9% PPV, a Kappa coefficient of 95.4%, and an F1-score of 95.5%.Keywords: These results highlight the model's effectiveness, adaptability and robustness for HSI classification. 2025 IEEE.

Source

Proceedings of 5th International Conference on Ubiquitous Computing and Intelligent Information Systems, ICUIS 2025;pp.151-157

Date

01-01-2025

Publisher

Institute of Electrical and Electronics Engineers Inc.

Subject

deep learning; HSI classification; Hyperspectral images; MNF; optimization

Coverage

Athul Ghosh P., Christ University, School of Engineering and Technology, Department of Ece, Bangalore, India; Chidambaram S., Christ University, School of Engineering and Technology, Department of Ece, Bangalore, India

Rights

Restricted Access; Hardcopy may be available in the library

Relation

ISBN: 979-833155284-8;

Format

online

Language

English

Type

Conference paper

Identifier

https://doi.org/10.1109/ICUIS67429.2025.11380606

https://www.scopus.com/pages/publications/105034832266?origin=resultslist