Spatial–Spectral Split Attention Residual Network for Hyperspectral Image Classification

Shu, Zhenqiu; Liu, Zigao; Zhou, Jun; Tang, Songze; Yu, Zhengtao; Wu, Xiao-Jun

doi:10.1109/jstars.2022.3225928

Cited by 17 publications

(9 citation statements)

References 62 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…After the feature extraction architecture, the global average pooling layer is first added to compress the feature map parameters to reduce its dimensions 37 . Next, the LN normalized feature distribution is accessed to facilitate the classification of tree species data, and finally, the fully connected layer with node 10 is accessed to discriminate the class of tree species.…”

Section: Methodsmentioning

confidence: 99%

Simple linear iterative clustering and ConvNeXt for mapping vectorize tree species

Wang

Chen

2023

J. Appl. Rem. Sens.

View full text Add to dashboard Cite

We propose a pioneering approach for gathering data on the forest canopy, one that merges two cutting-edge technologies: ConvNext tiny and simple linear iterative clustering (SLIC) (ConvNeXt and SLIC vectorize for tree species mapping, CSVTSM). By leveraging the clustering label generated by SLIC, CSVTSM obtains the vectorized result of land cover and allows us to obtain the location and distribution range information about the individual canopy. Moreover, CSVTSM employed ConvNeXt tiny, an advanced model, to identify vector objects and produce vector tree species maps across the experimental area to obtain the area of various tree species. To evaluate the accuracy of our methodology, we compared the canopy boundary of the original image to the four experimental areas that were vectorized by CSVTSM. Our model's efficacy was assessed by calculating the overall accuracy and Kappa coefficient on the validation set. The effectiveness of the proposed method was qualitatively evaluated by calculating the difference between the tree species area extracted by CSVTSM and manual extraction. The results indicate that, when considering the same canopy size, the vectorization approach based on the SLIC-based clustering label generates vector boundaries that are more closely aligned with the distribution of actual canopy boundaries in the experimental area. Furthermore, when used in conjunction with the SLIC vectorization approach, the ConvNeXt model can produce an even more precise vector map of tree species and more accurate tree species area information is obtained (only a 0.26 ha difference from manual extraction). These findings demonstrate that CSVTSM can be used to quantitatively acquire a wealth of information on individual tree positions, crown distribution ranges, and planting areas from high-resolution RGB photos captured by low-consumption unmanned aerial vehicle platforms. The implications of these results are wide-ranging, with local managers expressing keen interest in our findings as they offer trustworthy support for statistical application fields, such as gathering area ecological information, locating tree species resources, and disseminating spatial information about forests..

show abstract

Section: Methodsmentioning

confidence: 99%

Simple linear iterative clustering and ConvNeXt for mapping vectorize tree species

Wang

Chen

2023

J. Appl. Rem. Sens.

View full text Add to dashboard Cite

show abstract

“…In recent spectral-spatial methods, [17][18][19][20] CNNs are commonly employed to extract spectralspatial features from adjacent pixels, and convolution is an important component. [21][22][23] Recently, attention mechanisms were developed by simulating the visual system of humans, which selectively concentrates on prominent parts rather than handling each part consistently. 24 In HSI classification area, there are usually two types: spectral attention modules and spatial attention modules.…”

Section: Introductionmentioning

confidence: 99%

“…In recent spectral-spatial methods, 17 – 20 CNNs are commonly employed to extract spectral-spatial features from adjacent pixels, and convolution is an important component 21 – 23 …”

Section: Introductionmentioning

confidence: 99%

Center-similarity spectral-spatial attention network for hyperspectral image classification

Zhang,

Liang,

Niu

et al. 2024

J. Appl. Rem. Sens.

View full text Add to dashboard Cite

Hyperspectral image (HSI) classification aims to assign labels to pixels to be classified. The high-dimensional form of HSI and the introduction of spatial information can introduce challenges such as redundant spectral bands and interference pixels. Recently, many methods based on convolutional neural networks and attention mechanisms have been employed to address these issues. However, existing methods often fail to adequately utilize spatial information and exhibit interference pixel diffusion, which may result in these methods being unable to extract discriminative features. In addition, convolutions are not capable of extracting robust features, which results in the poor robustness of recent methods when HSI is rotated. To address these challenges, a center-similarity spectral-spatial attention network model (CS 3 AN) is proposed. First, a center-similarity spectral attention (CSSpeA) module is proposed to exploit the input spatial information rationally to better reduce the impact of redundant bands. Next, a center-similarity rectified spatial attention (CSRSpaA) module that selectively weighs neighboring pixels based on interference pixels to prevent the diffusion phenomenon is provided. Second, a similarity spatial aggregation module is employed to extract robust spectral-spatial features. Finally, the robust features are input into the softmax to get the label. The validity of the CS 3 AN is validated on three different databases.

show abstract

“…Although this approach is simple and compen-sates for the lack of spectral information, directly inputting the entire sample cube into the model results in a significant amount of redundant information interfering with feature extraction. Some studies have separated spatial and spectral information into different samples and used cross-domain contrastive learning to extract them separately [28][29][30]. This approach can reduce a lot of redundant information but may also lead to the loss of valuable sub-key information.…”

Section: Introductionmentioning

confidence: 99%

Unlocking the Potential of Data Augmentation in Contrastive Learning for Hyperspectral Image Classification

Yan

2023

Remote Sensing

View full text Add to dashboard Cite

Despite the rapid development of deep learning in hyperspectral image classification (HSIC), most models require a large amount of labeled data, which are both time-consuming and laborious to obtain. However, contrastive learning can extract spatial–spectral features from samples without labels, which helps to solve the above problem. Our focus is on optimizing the contrastive learning process and improving feature extraction from all samples. In this study, we propose the Unlocking-the-Potential-of-Data-Augmentation (UPDA) strategy, which involves adding superior data augmentation methods to enhance the representation of features extracted by contrastive learning. Specifically, we introduce three augmentation methods—band erasure, gradient mask, and random occlusion—to the Bootstrap-Your-Own-Latent (BYOL) structure. Our experimental results demonstrate that our method can effectively improve feature representation and thus improve classification accuracy. Additionally, we conduct ablation experiments to explore the effectiveness of different data augmentation methods.

show abstract

Spatial–Spectral Split Attention Residual Network for Hyperspectral Image Classification

Cited by 17 publications

References 62 publications

Simple linear iterative clustering and ConvNeXt for mapping vectorize tree species

Simple linear iterative clustering and ConvNeXt for mapping vectorize tree species

Center-similarity spectral-spatial attention network for hyperspectral image classification

Unlocking the Potential of Data Augmentation in Contrastive Learning for Hyperspectral Image Classification

Contact Info

Product

Resources

About