Title: Profound feature extraction and classification of hyperspectral images using deep learning

Authors: R. Venkatesan; Sevugan Prabu

Addresses: School of Computer Science and Engineering, Vellore Institute of Technology, Vellore, India ' School of Computer Science and Engineering, Vellore Institute of Technology, Vellore, India

Abstract: A hyperspectral image provides an asset of information, which tackles diversity of problems that occur in remote sensing the earth. These hyperspectral images are used in various applications such as research in geology, research in global changes, etc. The persistent issue, recurrent problem in remote sensing is classifying the spectral images, in which each pixels of the spectral image are being aimed to assign a label. Many steps are followed in hyper-spectral image classification such as pre-processing, features extraction with dimensionality reduction and classification. The pre-processing is used to reduce the noises using median filtering with anisotropic diffusion (AD) approach. Dimensionality reduction technique is often used to reduce the large volume of data. Discriminative local metric learning approach is applied to reduce the hyper spectral cube image dimension, thus retaining the significant components for further processing. Finally. recurrent neural network (RNN) algorithm is implemented to classify the features with improved accuracy. The proposed work outperforms the existing approaches in terms of accuracy and error rate measurements.

Keywords: hyper spectral imaging; diffusion approach; classification; features extraction; neural networks; class labels.

DOI: 10.1504/IJIE.2022.10046418

International Journal of Intelligent Enterprise, 2022 Vol.9 No.3, pp.318 - 331

Received: 24 Apr 2019
Accepted: 26 Sep 2019
Published online: 04 Jul 2022 *

Full-text access for editors Access for subscribers Purchase this article Comment on this article