To address the problems of insufficient utilization of spatial and spectral information and the need for large number of training samples in the current hyperspectral image classification methods based on deep learning, a algorithm(or model) named 3D Convolution Capsule Network (3D Conv-CapsNet) is investigated in this paper. A 3D convolution-capsule layer is constructed by combining the local connectivity in CNN and the capsule layer in Capsule Network. With the 3D convolution-capsule layer as the core, 3D Conv-CapsNet can acquire and utilize both spatial and spectral information of hyperspectral images, reduce the number of parameters and computational cost of the model, improve the feature representation capability of the capsule, and provide accurate classification results even in the case of limited training samples.In this paper, the classification results of 3D Conv-CapsNet on two hyperspectral image datasets, Salinas (SA) and Pavia University (PU), are analyzed and compared with SVM, 1D-CNN and 2D-CNN and 3D DenseNet. The experimental results show that, with 10% training samples, 3D Conv-CapsNet achieves 99.96%, 99.89% overall classification accuracy on SA and PU, which is better than SVM, 1D-CNN and 2D-CNN and 3D DenseNet. And the overall classification accuracy of 3D Conv-CapsNet network is still over 97% when the training samples are reduced to 5%, 3% and 1%.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.