Shu Xiang

Dr. Shu Xiang

orcid

Research Focus & Metrics

Academic Career Length

9Years (2018-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
19 December 2023 in IEEE Transactions on Geoscience and Remote Sensing
Verified
Journal Article
Closed access
JDJigang Ding
Jigang Ding
XLXiaorun Li
Xiaorun Li
SXShu Xiang
Shu Xiang
Shuhan ChenShuhan Chen
Shuhan Chen
Hyperspectral image change detection (HSI CD) refers to identifying and analyzing differences between two HSIs acquired in the same area but at different times. However, current deep learning-based methods have limitations in fully exploiting the change information between bi-temporal images and utilizing multi-level features. To address these issues, we propose a novel Multi-level features Fused and Change information Enhanced neural Network (MFCEN) for HSI CD. The proposed MFCEN method leverages the hierarchical low- and high-level features of bi-temporal images, allowing for the direct capture and enhancement of change features. First, A Siamese-based network is employed to extract multi-level features from the bi-temporal images, including low-level spatial details and high-level semantic features. Within the temporal change information branch (TCIB) at each level, the change features are reinforced by the semantic features of each image, and the change features act as a guiding force to direct the feature extraction of each bi-temporal image to focus more on the change region. Next, the enhanced change features of each level are fed into the multi-level features fusion module (MFFM) to aggregate the fine-grained details and high-level semantics. Finally, the fused features, enriched with multi-level change information, are utilized for CD. The experiments demonstrate that our proposed MFCEN outperforms existing methods on three public datasets. The code will be available at https://github.com/Ding201901/MFCEN.

16 January 2025 in IEEE Transactions on Geoscience and Remote Sensing
Verified
Journal Article
Closed access
SXShu Xiang
Shu Xiang
XLXiaorun Li
Xiaorun Li
Shuhan ChenShuhan Chen
Shuhan Chen
In recent years, researchers have focused on mitigating the impact of spectral variability (SV) on unmixing performance, leading to the development of various deep learning-based unmixing networks. Currently, most unmixing networks that account for SV mainly rely on probabilistic generative models, which lacks specific constraints for SV and suffers from the instability of solutions generated by probabilistic models. This results in the generated endmembers or SV components lacking clear physical meaning. To avoid the problems above, we propose an endmember-oriented Transformer network (EOT-Net) that leverages the advantages of endmember bundles to introduce variability while providing stable endmember results with clear physical meaning. We design an endmember-oriented Transformer (EOT) to capture endmember-specific features through directional subspace projection and a low-redundancy attention mechanism. Subsequently, the proposed network is divided into two branches: endmember generation and abundance estimation, to process endmember-specific features. In the endmember generation branch, endmember-specific features are transformed into intra-class weights that are used to combine signatures within the bundles, and a set of endmembers is generated for each pixel. In the abundance estimation branch, endmember-specific features are integrated using a heterogeneous information fusion (HIF) module that leverages the spatial distribution heterogeneity of the endmembers, ultimately producing the abundance results. We applied the proposed algorithm to both synthetic and real datasets, and the experimental results demonstrated the model’s superiority.

14 March 2024 in IEEE Transactions on Geoscience and Remote Sensing
Verified
Journal Article
Closed access
SXShu Xiang
Shu Xiang
XLXiaorun Li
Xiaorun Li
JDJigang Ding
Jigang Ding
Shuhan ChenShuhan Chen
Shuhan Chen
ZHZiqiang Hua
Ziqiang Hua
Autoencoders (AEs) has demonstrated excellent performance in the field of hyperspectral unmixing, due to their self-supervised nature and ease of implementation. Recently proposed AE-based networks contend that local spatial information limits further improvement in unmixing accuracy and tend to explore and utilize global information, which improves unmixing accuracy at the expense of increased computational complexity. However, we believe that precise unmixing can be achieved by fully leveraging local information. In this paper, we propose a unidirectional local-attention autoencoder Network (ULA-Net) that explores spatial information pixel by pixel and achieves accurate spatial-spectral feature fusion. ULA-Net utilizes unidirectional local attention (ULA) module to calculate the correlation between neighboring pixels and the central pixel within local regions, extracting discriminative local information. Moreover, ULA-Net effectively extracts relevant spatial information and suppresses irrelevant information based on a double fusion strategy (DFS) module. This process achieves more accurate control over the contribution of spatial information by implementing information fusion in both the pixel and feature dimensions. To address spectral variability, we implement the extended linear mixing model (ELMM) in the decoder part to improve unmixing accuracy without increasing the number of parameters. We conduct ablation experiments to investigate the roles of each module. Experimental results on both synthetic and real datasets demonstrate the effectiveness of the proposed network.

30 August 2024 in IEEE Transactions on Geoscience and Remote Sensing
Verified
Journal Article
Closed access
SXShu Xiang
Shu Xiang
XLXiaorun Li
Xiaorun Li
Shuhan ChenShuhan Chen
Shuhan Chen
Hyperspectral images encounter an inherent challenge due to spectral variability (SV), which directly impacts the accuracy of endmember extraction and abundance estimation. Most unmixing networks based on autoencoder (AE) overlook the impact of SV and instead focus more on exploring the features of endmembers. In this paper, we propose a feedback information-guided spectral variability attention network (FSVA-Net), an AE-based neural network that utilizes a feedback information-guided structure to exploit and model the SV factors. We develop a feedback enhancement module that utilizes residuals from a linearly reconstructed image to reweight and emphasize pixels affected by spectral variability. And a spectral variability attention (SVA) mechanism is proposed to explore interference-affected information in a global perspective according to the enhanced feature. With the extracted SV-related feature, we design a generative augmented linear mixing model (ALMM)-based decoder to model SV from the perspective of scaling and perturbation in a more reliable way. In a more concern on SV architecture, the proposed network achieves excellent performance on both synthetic and real datasets.

2 March 2026 in IEEE Transactions on Geoscience and Remote Sensing
Verified
Journal Article
Closed access
JSJiahui Sheng
Jiahui Sheng
SXShu Xiang
Shu Xiang
XLXiaorun Li
Xiaorun Li
Shuhan ChenShuhan Chen
Shuhan Chen
Hyperspectral images (HSIs), as a unique type of 3-D image, have a wide range of applications across various fields. However, their multichannel characteristics make them challenging for direct visual observation by humans, leading to significant inconveniences. To address this issue, the most common approach is to select three channels to generate an RGB image for visualization. However, this method neglects the most crucial spectral information in HSI, failing to represent the spectral composition of objects. Therefore, in this study, we proposed an HSI visualization method called SaHV, which aims to represent the spectral composition in the visualization results. The specific process of SaHV is as follows. First, the proposed method used spectral unmixing to obtain the abundance maps, which stands for the spectral composition. Then the obtained abundance maps were processed by our color mapping network (CMN) to produce hue and saturation channels of HSV, in which the force-directed layout method was introduced to keep local consistency. Finally, the original HSI was used to obtain the value channel of HSV. To demonstrate the effectiveness of our method, we conducted experiments on three real-world datasets.

1 September 2018
Verified
Conference Paper
Closed access
CHCong Haoxi
Cong Haoxi
SXShu Xiang
Shu Xiang
ZMZhang Minhao
Zhang Minhao
RPRen Peng
Ren Peng
DSDu Shiyue
Du Shiyue
JHJin Hu
Jin Hu
The corrosive sulfur phenomenon has decades of history and currently adding passivators to the transformer oil is the most effective way to suppress the corrosive sulfur problem. However, it has not been reported how to choose the passivator concentration with different kinds of corrosive sulfur or different aging degree. In this paper, dibenzyl disulfide (DBDS) and dodecyl mercaptan (DDM) were added to the non-corrosive 25# transformer new oil, also with different concentrations of Irgamet 39, T571 and TTA. The accelerated thermal aging test was carried out at 150 °C temperature, and the corrosive extent of copper surface were observed to obtain the best concentration of passivators for different sulfurs. The accelerated thermal aging test were also carried out at a temperature of 130 °C for different aging degree. The passivators were added to the oil and the corrosion extent of copper surface were observed to get the best passivator concentration for oil with different aging degree. Results show that three passivators have different protective effects for different corrosive sulfurs. Irgamet 39 has a better protective effect than others at low concentration. However, both three kinds of passivators have poor protective effect on dodecyl mercaptan, and a new kind of passivator need to be developed. For different aging degree oil, the passivators also have different protective effect. For the transformer oil of higher aging degree, the passivators in low concentration have poor protective effect, thus the passivator content need to be concerned to make it sure at high concentration level. The research work can provide some reference for guiding the addition of passivator in practical engineering and preventing the occurrence of sulfur corrosion failure.