Going Deeper With Contextual CNN for Hyperspectral Image Classification

In this paper, we describe a novel deep convolutional neural network (CNN) that is deeper and wider than other existing deep networks for hyperspectral image classification. Unlike current state-of-the-art approaches in CNN-based hyperspectral image classification, the proposed network, called conte...

Full description

Saved in:

Bibliographic Details
Published in:	IEEE transactions on image processing 2017-10, Vol.26 (10), p.4843-4855
Main Authors:	Lee, Hyungtae, Kwon, Heesung
Format:	Article
Language:	eng
Subjects:	Artificial neural networks Biological neural networks Classification Convolutional neural network (CNN) Datasets Feature extraction Feature maps fully convolutional network (FCN) hyperspectral image classification Hyperspectral imaging Image classification multi-scale filter bank Neural networks Pixels Principal component analysis residual learning Spectra State of the art Training
Online Access:	Get full text
Tags:	Add Tag No Tags, Be the first to tag this record!

Description
Summary:	In this paper, we describe a novel deep convolutional neural network (CNN) that is deeper and wider than other existing deep networks for hyperspectral image classification. Unlike current state-of-the-art approaches in CNN-based hyperspectral image classification, the proposed network, called contextual deep CNN, can optimally explore local contextual interactions by jointly exploiting local spatio-spectral relationships of neighboring individual pixel vectors. The joint exploitation of the spatio-spectral information is achieved by a multi-scale convolutional filter bank used as an initial component of the proposed CNN pipeline. The initial spatial and spectral feature maps obtained from the multi-scale filter bank are then combined together to form a joint spatio-spectral feature map. The joint feature map representing rich spectral and spatial properties of the hyperspectral image is then fed through a fully convolutional network that eventually predicts the corresponding label of each pixel vector. The proposed approach is tested on three benchmark data sets: the Indian Pines data set, the Salinas data set, and the University of Pavia data set. Performance comparison shows enhanced classification performance of the proposed approach over the current state-of-the-art on the three data sets.
ISSN:	1057-7149 1941-0042