Multi-Stage Convolutional Broad Learning with Block Diagonal Constraint for Hyperspectral Image Classification

By combining the broad learning and a convolutional neural network (CNN), a block-diagonal constrained multi-stage convolutional broad learning (MSCBL-BD) method is proposed for hyperspectral image (HSI) classification. Firstly, as the linear sparse feature extracted by the conventional broad learni...

Full description

Saved in:
Bibliographic Details
Published inRemote sensing (Basel, Switzerland) Vol. 13; no. 17; p. 3412
Main Authors Kong, Yi, Wang, Xuesong, Cheng, Yuhu, Chen, C. L. Philip
Format Journal Article
LanguageEnglish
Published Basel MDPI AG 01.09.2021
Subjects
Approximation
Artificial neural networks
block diagonal
broad learning system
Classification
Constraints
convolutional neural network
Deep learning
Feature extraction
Feature selection
hyperspectral image
Hyperspectral imaging
Image classification
Information processing
Learning
Mapping
Methods
Multilayers
Neural networks
Principal components analysis
Remote sensing
Online AccessGet full text

Cover

More Information
Summary:By combining the broad learning and a convolutional neural network (CNN), a block-diagonal constrained multi-stage convolutional broad learning (MSCBL-BD) method is proposed for hyperspectral image (HSI) classification. Firstly, as the linear sparse feature extracted by the conventional broad learning method cannot fully characterize the complex spatial-spectral features of HSIs, we replace the linear sparse features in the mapped feature (MF) with the features extracted by the CNN to achieve more complex nonlinear mapping. Then, in the multi-layer mapping process of the CNN, information loss occurs to a certain degree. To this end, the multi-stage convolutional features (MSCFs) extracted by the CNN are expanded to obtain the multi-stage broad features (MSBFs). MSCFs and MSBFs are further spliced to obtain multi-stage convolutional broad features (MSCBFs). Additionally, in order to enhance the mutual independence between MSCBFs, a block diagonal constraint is introduced, and MSCBFs are mapped by a block diagonal matrix, so that each feature is represented linearly only by features of the same stage. Finally, the output layer weights of MSCBL-BD and the desired block-diagonal matrix are solved by the alternating direction method of multipliers. Experimental results on three popular HSI datasets demonstrate the superiority of MSCBL-BD.
ISSN:2072-4292
2072-4292
DOI:10.3390/rs13173412