HYPERSPECTRAL IMAGE CLASSIFICATION ON A MULTISCALE 3D CONVOLUTIONAL NEURAL NETWORK

HYPERSPECTRAL IMAGE CLASSIFICATION ON A MULTISCALE 3D CONVOLUTIONAL NEURAL NETWORK

In the past decade, significant progress has been made in hyperspectral image (HSI) classification due to advances in deep learning techniques. This study introduces a novel computational framework for classifying HSIs using a multiscale 3D convolutional neural network (3DCNN). The purpose of this r...

Full beskrivning

Sparad:
Bibliografiska uppgifter
Huvudupphovsmän: Thai Duy Quy, Tran, Thi Ngoc Thao, Avirmed, Enkhbat
Materialtyp: Journal article
Språk:English
Publicerad: 2025
Ämnen:
2DCNN
3DCNN
Deep learning
HSI datasets
Hyperspectral image
Recurrent neural networks
Temporal convolutional networks.
Länkar:https://scholar.dlu.edu.vn/handle/123456789/5801
https://tckh.dlu.edu.vn/index.php/tckhdhdl/article/view/1381
Taggar: Lägg till en tagg
Inga taggar, Lägg till första taggen!
Thư viện lưu trữ: Thư viện Trường Đại học Đà Lạt
Beskrivning
Sammanfattning:In the past decade, significant progress has been made in hyperspectral image (HSI) classification due to advances in deep learning techniques. This study introduces a novel computational framework for classifying HSIs using a multiscale 3D convolutional neural network (3DCNN). The purpose of this research is to enhance the efficiency of HSI classification by leveraging the spatial-spectral characteristics of HSI data. The methodology involves preprocessing HSI datasets, extracting features using parallel 3DCNN layers with skip connections, and evaluating performance on five benchmark datasets. Key findings demonstrate that the proposed model significantly improves classification accuracy, achieving average accuracies of 85%, 84%, 94%, 98%, and 95% on the Kennedy Space Center (KSC), Indian Pines (IP), Pavia University (PU), Pavia Centre (PC), and Botswana (BW) datasets, respectively. The originality of this study lies in the multiscale parallel design of the 3DCNN architecture, which balances computational complexity with enhanced performance. Limitations include the high computational cost associated with training deep networks, which can be mitigated by graphics processing unit (GPU) acceleration and optimized implementations.

Liknande verk