AAtt-CNN: Automatic Attention-Based Convolutional Neural Networks for Hyperspectral Image Classification

Paoletti, Mercedes E.; Moreno‐Álvarez, Sergio; Xue, Yu; Haut, Juan M.

doi:10.1109/tgrs.2023.3272639

Cited by 27 publications

(7 citation statements)

References 82 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Involving recalibration of the spectrum features. Numerous scholars have successfully integrated attention mechanisms into spectral similarity measurements, effectively aggregating similar spectra and optimize cognitive performance of convolutional neural networks [44], [45], [46], [47].…”

Section: A Prerequisitementioning

confidence: 99%

PCCN-MSS: Parallel Convolutional Classification Network Combined Multi-Spatial Scale and Spectral Features for UAV-Borne Hyperspectral With High Spatial Resolution Imagery

Jiang,

Zhang,

Tang

et al. 2024

IEEE J. Sel. Top. Appl. Earth Observations Remote Sensing

View full text Add to dashboard Cite

Hyperspectral remote sensing images with high spatial resolution (H 2 imagery) have an abundant spatial-spectral information, holding tremendous potential for remote sensing fine-grained monitoring and classification. However, challenges such as high spatial heterogeneity, severe intra-class spectral variability and poor signal-to-noise ratio especially in unmanned aerial vehicle (UAV) hyperspectral imagery constrain and hinder the performance of fine-grained classification. Convolutional neural network (CNN) emerges as a formidable and excellent tool for image mining and feature extraction, offering effective utility for land cover classification. In this paper, a parallel convolutional classification network model based on multi-modal filters (including ICA-2D-FPN and SA-3D-CNN branching structures) PCCN-MSS is proposed for precise H 2 imagery classification. The ICA-2D-FPN branch integrates independent component analysis (ICA) into 2D-CNN to extract the multispatial scale and spectral information of H 2 imagery by Feature Pyramid Networks (FPN), meanwhile the SA-3D-CNN branch is designed to extract the spatial and spectral information by combining spectral attention (SA) mechanism and 3D-CNN. Taking hyperspectral imagery of unmanned aerial vehicles containing vegetation and artifactual material ground as an example, the proposed PCCN-MSS model achieves an overall accuracy (OA) of 78.18%, which outperforms by 9.58% than the compared methods. The proposed PCCN-MSS method can mitigate the classification issues of severe salt-and-pepper noise and inaccurate boundary, delivering more satisfactory classification results with robust classification performance and remarkable advantages for H 2 imagery.

show abstract

Section: A Prerequisitementioning

confidence: 99%

PCCN-MSS: Parallel Convolutional Classification Network Combined Multi-Spatial Scale and Spectral Features for UAV-Borne Hyperspectral With High Spatial Resolution Imagery

Jiang,

Zhang,

Tang

et al. 2024

IEEE J. Sel. Top. Appl. Earth Observations Remote Sensing

View full text Add to dashboard Cite

show abstract

“…The depth of a neural network will affect its capacity. The term deep learning (or deep neural networks) refers to neural networks with at least one hidden layer [19].…”

Section: Neural Networkmentioning

confidence: 99%

“…Furthermore, more complex and better performance methods also exist to optimise the weights, such as the Levenberg-Marquardt algorithm [32]. At present, deep networks and their ensembles are the most advanced solutions for the majority of typical applications, including computer vision, speech processing, and image processing [14][15][16][17][18][19]. However, such solutions rely on "heavy" models and a large number of parameters, which can potentially be computationally expensive and hamper their use in real-time systems.…”

Section: Neural Networkmentioning

confidence: 99%

“…In fact, the time delays in SAS can be understood as a mapping relationship between the echoes and the platform motion, which suggests that supervised learning techniques may be effectively utilised in the time delay estimation of SAS. There are many different forms of mapping in supervised learning, including decision trees, logistic regression, kernel machines, support vector machines (SVMs), Bayesian classifiers, and neural networks [14][15][16][17][18][19][20][21][22][23][24][25][26]. In recent years, deep neural networks have made remarkable progress in the field of supervised learning [14].…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Deep learning‐based time delay estimation for motion compensation in synthetic aperture sonars

Chen,

Chi,

Zhang

et al. 2023

IET Radar Sonar & Navi

View full text Add to dashboard Cite

Accurate and robust time delay estimation is crucial for synthetic aperture sonar (SAS) imaging. A two‐step time delay estimation method based on displaced phase centre antenna (DPCA) micronavigation has been widely applied in motion estimation and compensation of SASs. However, the existing methods for time delay estimation are not sufficiently robust, which reduces the performance of SAS motion estimation. Deep learning is currently one of the cutting‐edge techniques and has brought about a remarkable progress in the field of underwater acoustic signal processing. In this study, a deep learning‐based time delay estimation method is introduced to SAS motion estimation and compensation. The subband processing is first applied to obtain ambiguous time delays between adjacent pings from phases of SAS echoes. Then, a lightweight neural network is utilised to construct phase unwrapping. The model of the employed neural network is trained with simulation data and applied to real SAS data. The results of time delay estimation and motion compensation demonstrate that the proposed neural network‐based method has much better performance than the two‐step and joint‐subband methods.

show abstract

“…Zhao et al [37] designed channel and spatial residual attention modules to integrate the multi-scale features by introducing a residual connection and attention mechanism. Paoletti et al [38] employed channel attention to automatically design a network to extract spatial and spectral features. Shi et al [39] proposed the 3D-OCONV method and abstracted the spectral features by combining 3D-OCONV and spectral attention.…”

Section: Attention Mechanismmentioning

confidence: 99%

A Dual-Attention Deep Discriminative Domain Generalization Model for Hyperspectral Image Classification

Zhao,

Wang,

Wang

et al. 2023

Remote Sensing

View full text Add to dashboard Cite

Recently, hyperspectral image classification has made great progress with the development of convolutional neural networks. However, due to the challenges of distribution shifts and data redundancies, the classification accuracy is low. Some existing domain adaptation methods try to mitigate the distribution shifts by training source samples and some labeled target samples. However, in practice, labeled target domain samples are difficult or even impossible to obtain. To solve the above challenges, we propose a novel dual-attention deep discriminative domain generalization framework (DAD3GM) for cross-scene hyperspectral image classification without training the labeled target samples. In DAD3GM, we mainly design two blocks: dual-attention feature learning (DAFL) and deep discriminative feature learning (DDFL). DAFL is designed to extract spatial features by multi-scale self-attention and extract spectral features by multi-head external attention. DDFL is further designed to extract deep discriminative features by contrastive regularization and class discrimination regularization. The combination of DAFL and DDFL can effectively reduce the computational time and improve the generalization performance of DAD3GM. The proposed model achieves 84.25%, 83.53%, and 80.63% overall accuracy on the public Houston, Pavia, and GID benchmarks, respectively. Compared with some classical and state-of-the-art methods, the proposed model achieves optimal results, which reveals its effectiveness and feasibility.

show abstract

AAtt-CNN: Automatic Attention-Based Convolutional Neural Networks for Hyperspectral Image Classification

Cited by 27 publications

References 82 publications

PCCN-MSS: Parallel Convolutional Classification Network Combined Multi-Spatial Scale and Spectral Features for UAV-Borne Hyperspectral With High Spatial Resolution Imagery

PCCN-MSS: Parallel Convolutional Classification Network Combined Multi-Spatial Scale and Spectral Features for UAV-Borne Hyperspectral With High Spatial Resolution Imagery

Deep learning‐based time delay estimation for motion compensation in synthetic aperture sonars

A Dual-Attention Deep Discriminative Domain Generalization Model for Hyperspectral Image Classification

Contact Info

Product

Resources

About