Hyperspectral Image Classification With Deep Feature Fusion Network

By Weiwei Song, Shutao Li, Leyuan Fang, Ting Lu. IEEE Transactions on Geoscience and Remote Sensing, 2018

450 citations recorded by the source corpus.

Recently, deep learning has been introduced to classify hyperspectral images (HSIs) and achieved good performance. In general, deep models adopt a large number of hierarchical layers to extract features. However, excessively increasing network depth will result in some negative effects (e.g., overfitting, gradient vanishing, and accuracy degrading) for conventional convolutional neural networks. In addition, the previous networks used in HSI classification do not consider the strong complementary yet correlated information among different hierarchical layers. To address the above two issues, a deep feature fusion network (DFFN) is proposed for HSI classification. On the one hand, the residual learning is introduced to optimize several convolutional layers as the identity mapping, which can ease the training of deep network and benefit from increasing depth. As a result, we can build a very deep network to extract more discriminative features of HSIs. On the other hand, the proposed DFFN model fuses the outputs of different hierarchical layers, which can further improve the classification accuracy. Experimental results on three real HSIs demonstrate that the proposed method outperforms other competitive classifiers.

Subfields: Atmospheric Science, Media Technology

View Paper: https://www.semanticscholar.org/paper/1df49ddb9460f302ff9692f4b93c9f3d68693a17

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Atmospheric Science

In Media Technology

Citations in the corpus

Attention-Based Adaptive Spectral–Spatial Kernel ResNet for Hyperspectral Image Classification (2020, methodology): “The layer-wise feature fusion without attention [21], the multibranch selective kernel (SK) network with attention [22], and the data augmentation strategy [23], [24] were employed to deal with gradient vanishing and overfitting problems.”
Feature Extraction for Hyperspectral Imagery: The Evolution From Shallow to Deep: Overview and Toolbox (2020, methodology): “In [100], [101], [103], PCA was employed to extract the first principal components of hyperspectral images as inputs of CNNs, thus simplifying the network structures.”
HybridSN: Exploring 3-D–2-D CNN Feature Hierarchy for Hyperspectral Image Classification (2019, background): “[23] have proposed a deep feature fusion network and Cheng et al.”
Multiscale Dynamic Graph Convolutional Network for Hyperspectral Image Classification (2019, background): “In contrast, deep learning-based methods [24], [48], [49], which can generate features automatically, have recently attracted increasing attention in hyperspectral image classification.”
Deep Learning Meets Hyperspectral Image Analysis: A Multidisciplinary Review (2019, methodology): “In doing so, we will introduce various methods, from classical networks to the integration with multiscale and fusion strategies, as in [131].”
Deep Learning for Hyperspectral Image Classification: An Overview (2019, methodology): “The main drawback of DFFN is that the optimal feature fusion mechanism depends on a hand-crafted setting with abundant experiments.”
A Spectral-Spatial-Dependent Global Learning Framework for Insufficient and Imbalanced Hyperspectral Image Classification (2021, background): “[16] proposed RNNs extracting the spectral dependency among adjacent wavebands and regarded the HSI”