Remote Sensing Image Scene Classification Based on Global Self-Attention Module

Li, Qingwen; Yan, Dongmei; Wu, Wanrong

doi:10.3390/rs13224542

Cited by 12 publications

(5 citation statements)

References 58 publications

(61 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Most studies use the hyperbola fitting method due to its simplicity and computational efficiency (Dong et al 2020a,b;Lai et al 2019;Song et al 2021). Giannakis et al (2021) improved the method by considering the permittivity as varying with depth, while Li & Zhang (2021) used the migration method and separated the diffraction from reflections to estimate the permittivity.…”

Section: Introductionmentioning

confidence: 99%

“…The derivation of permittivity distribution along the horizontal direction is also limited by the absence of a hyperbola shaped signal. The permittivity map obtained by (Dong et al 2020b;Li & Zhang 2021;Song et al 2021) was produced by interpolation based on a couple of sparse points. The resolution is too coarse to reveal the finer details, which are necessary for a reliable mapping of the geological stratigraphy.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Sub-surface stratification and dielectric permittivity distribution at the Chang’E-4 landing site revealed by the lunar penetrating radar

Chen

Xie

et al. 2022

A&A

View full text Add to dashboard Cite

Context. In 2019, China’s Chang’E-4 (CE-4) probe landed on the far side of the Moon: a first in lunar exploration. The Lunar Penetrating Radar (LPR) mounted on the Yutu-2 rover allows the mapping of the near-surface structure and the dielectric permittivity of the landing area. The dielectric properties of the lunar soil affect the propagation of the LPR signals, which can be used to infer the depth of sub-surface boundaries and derive the composition of the component materials. Aims. Our objectives are to estimate the fine-resolution spatial distribution of relative permittivity and to improve the interpretation of the geological processes combined with the radargram of the CE-4 landing area. Methods. We used a modified method that combines the F-K migration and the minimum entropy of the ground penetrating radar (GPR) signals to estimate the velocity and permittivity values; this has the advantage of obtaining the appropriate velocity and permittivity, even with the incomplete or unnoticeable hyperbolic curves in the radar image Results. The sub-surface stratification of the CE-4 landing area is seen in the first 31 lunar days of the LPR data. A fine-resolution dielectric permittivity profile ranging from ~2.3 to ~6.3 is obtained with our method, and the actual depths of the observed prominent sub-surface interfaces are determined, giving a maximum average depth of ~38 m. The thickness of the regolith layer is in the range of ~5.7–15.6 m, with an average of 11.8 m. The permittivity of the near-surface regolith (<30 cm) is ~2.78 ± 0.01, the bulk density is 1.57 ± 0.01 g cm−3, which is close to the results of ~1.61 g cm−3 at the Apollo 15 landing area. The permittivity map is consistent with the radargram; the regolith and the paleo-regolith layer have relatively low permittivity and low echo strengths, while the rock debris has high permittivity and shows strong echos in the radargram. Two buried craters of different diameters beneath the navigation sites 4–11 and 16–31 are revealed in the radar profile. The permittivity distribution map can show detailed variations of material properties both inside and outside craters.

show abstract

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Sub-surface stratification and dielectric permittivity distribution at the Chang’E-4 landing site revealed by the lunar penetrating radar

Chen

Xie

et al. 2022

A&A

View full text Add to dashboard Cite

show abstract

“…A bundle of deep CNNs with diverse layouts of convolutional layers, pooling layers, and fully connected layers have been employed in scene classification, such as Visual Geometry Group (VGG)Net [12], Residual(Res)Net [13], and Densely Connected (Dense)Net [14], since AlexNet [15] obtained astonishing success. In recent years, CNN-based architectures have been used to improve classification accuracy [16][17][18][19][20][21][22]. For example, a dual-model architecture by combining two CNN-based branches, ResNet and DenseNet, with a globalattention-fusion strategy was designed in a recent study [23], which was proven effective by gaining higher accuracies than single-model architectures.…”

Section: Introductionmentioning

confidence: 99%

Multi-Output Network Combining GNN and CNN for Remote Sensing Scene Classification

Peng

Tan

et al. 2022

Remote Sensing

View full text Add to dashboard Cite

Scene classification is an active research area in the remote sensing (RS) domain. Some categories of RS scenes, such as medium residential and dense residential scenes, would contain the same type of geographical objects but have various spatial distributions among these objects. The adjacency and disjointness relationships among geographical objects are normally neglected by existing RS scene classification methods using convolutional neural networks (CNNs). In this study, a multi-output network (MopNet) combining a graph neural network (GNN) and a CNN is proposed for RS scene classification with a joint loss. In a candidate RS image for scene classification, superpixel regions are constructed through image segmentation and are represented as graph nodes, while graph edges between nodes are created according to the spatial adjacency among corresponding superpixel regions. A training strategy of a jointly learning CNN and GNN is adopted in the MopNet. Through the message propagation mechanism of MopNet, spatial and topological relationships imbedded in the edges of graphs are employed. The parameters of the CNN and GNN in MopNet are updated simultaneously with the guidance of a joint loss via the backpropagation mechanism. Experimental results on the OPTIMAL-31 and aerial image dataset (AID) datasets show that the proposed MopNet combining a graph convolutional network (GCN) or graph attention network (GAT) and ResNet50 achieves state-of-the-art accuracy. The overall accuracy obtained on OPTIMAL-31 is 96.06% and those on AID are 95.53% and 97.11% under training ratios of 20% and 50%, respectively. Spatial and topological relationships imbedded in RS images are helpful for improving the performance of scene classification.

show abstract

“…More recently, algorithms based on deep learning have achieved satisfactory results in many application fields owing to their strong feature representation power and powerful graphical processing units. Among the deep learning methods, convolutional neural networks (CNNs) have been successfully applied in remote sensing image scene classification [7][8][9][10][11][12][13][14]. Compared with traditional unsupervised feature learning methods, feature representations with more complex patterns can be learned via deep architecture neural networks.…”

Section: Introductionmentioning

confidence: 99%

Prob-POS: A Framework for Improving Visual Explanations from Convolutional Neural Networks for Remote Sensing Image Classification

Guo

Hou

et al. 2022

Remote Sensing

View full text Add to dashboard Cite

During the past decades, convolutional neural network (CNN)-based models have achieved notable success in remote sensing image classification due to their powerful feature representation ability. However, the lack of explainability during the decision-making process is a common criticism of these high-capacity networks. Local explanation methods that provide visual saliency maps have attracted increasing attention as a means to surmount the barrier of explainability. However, the vast majority of research is conducted on the last convolutional layer, where the salient regions are unintelligible for partial remote sensing images, especially scenes that contain plentiful small targets or are similar to the texture image. To address these issues, we propose a novel framework called Prob-POS, which consists of the class-activation map based on the probe network (Prob-CAM) and the weighted probability of occlusion (wPO) selection strategy. The proposed probe network is a simple but effective architecture to generate elaborate explanation maps and can be applied to any layer of CNNs. The wPO is a quantified metric to evaluate the explanation effectiveness of each layer for different categories to automatically pick out the optimal explanation layer. Variational weights are taken into account to highlight the high-scoring regions in the explanation map. Experimental results on two publicly available datasets and three prevalent networks demonstrate that Prob-POS improves the faithfulness and explainability of CNNs on remote sensing images.

show abstract

Remote Sensing Image Scene Classification Based on Global Self-Attention Module

Cited by 12 publications

References 58 publications

Sub-surface stratification and dielectric permittivity distribution at the Chang’E-4 landing site revealed by the lunar penetrating radar

Sub-surface stratification and dielectric permittivity distribution at the Chang’E-4 landing site revealed by the lunar penetrating radar

Multi-Output Network Combining GNN and CNN for Remote Sensing Scene Classification

Prob-POS: A Framework for Improving Visual Explanations from Convolutional Neural Networks for Remote Sensing Image Classification

Contact Info

Product

Resources

About