Hyperspectral Image Classification With Stacking Spectral Patches and Convolutional Neural Networks
- 16 May 2018
- journal article
- research article
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Geoscience and Remote Sensing
- Vol. 56 (10) , 5975-5984
Abstract
Jointly combining the spatial and spectral features has proved to dramatically improve the performance of classifying hyperspectral image. Recently, utilizing neural networks to automatically model the spatial-spectral feature representations for hyperspectral images has become of great interest. This paper proposes a simple but innovative framework to classify hyperspectral image with two shallow convolutional neural networks (CNNs). First, principal component analysis whitening is applied to decorrelate hundreds of spectral bands. Instead of selecting the principal components to reduce the spectral dimensionality, we retain all the spectral bands but compress the image cuboid into a one-channel spectral quilt by stacking spectral patches. In this way, not only all the spectral information is retained but also the computational complexity of training a neural network is reduced compared with conventional networks that directly input the spectral volumes. Moreover, the spectral quilt will contain some novel textural patterns that are effective at distinguishing classes. Two shallow CNNs are then applied to classify the spectral quilts. As shown in the experiments, both networks can outperform the standard analysis methods.Keywords
This publication has 33 references indexed in Scilit:
- A Survey on Spectral–Spatial Classification Techniques Based on Attribute ProfilesIEEE Transactions on Geoscience and Remote Sensing, 2014
- Extreme Learning Machine With Composite Kernels for Hyperspectral Image ClassificationIEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2014
- Supervised Spectral–Spatial Hyperspectral Image Classification With Weighted Markov Random FieldsIEEE Transactions on Geoscience and Remote Sensing, 2014
- Extended profiles with morphological attribute filters for the analysis of hyperspectral dataInternational Journal of Remote Sensing, 2010
- Hyperspectral Remote Sensing of VegetationGeography Compass, 2008
- The MARTE VNIR Imaging Spectrometer Experiment: Design and AnalysisAstrobiology, 2008
- A Fast Learning Algorithm for Deep Belief NetsNeural Computation, 2006
- Advances in mathematical morphology applied to geoscience and remote sensingIEEE Transactions on Geoscience and Remote Sensing, 2002
- A new approach for the morphological segmentation of high-resolution satellite imageryIEEE Transactions on Geoscience and Remote Sensing, 2001
- Thirteen Ways to Look at the Correlation CoefficientThe American Statistician, 1988