Xuedong Yao
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Xuedong Yao has a total of 17 co-authors. Here are the top co-authors by number of shared publications:
- Yandong Wang
State Key Laboratory of Information Engineering in Surveying, Mapping and Remote Sensing, Wuhan University, Wuhan, China
4 shared publications Lei Dai
State Key Laboratory of Information Engineering in Surveying, Mapping and Remote Sensing, Wuhan University, Wuhan, China
2 shared publicationsYanlan Wu
Information Materials and Intelligent Sensing Laboratory of Anhui Province, Anhui University, Hefei, China
2 shared publications- Zeyu Liang
Chinese Antarctic Center of Surveying and Mapping, Wuhan University, Wuhan 430079, China
2 shared publications Dongyang Wang
State Key Laboratory of Information Engineering in Surveying, Mapping and Remote Sensing, Wuhan University, Wuhan, China
1 shared publicationGuoxiong He
State Key Laboratory of Information Engineering in Surveying, Mapping and Remote Sensing (LIESMARS), Wuhan University, Wuhan, China
1 shared publication
Latest Publications
- Open Access
- 30 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
- Vol. 19 (19391404) , 16333-16345
Controllable Remote Sensing Image Generation With Semantic Consistency via Diffusion Models
The generation of remote sensing images plays a vital role in applications such as urban planning, urban development simulation, and environmental monitoring. The advent of diffusion models has introduced a new paradigm in image generation and achieved remarkable results across various fields. However, their adoption in remote sensing remains limited, primarily due to the lack of automated pipelines for constructing multimodal remote sensing datasets and the challenge of preserving semantic controllability given the complex object categories and high semantic similarity typical of remote sensing scenes. In this paper, we propose a remote sensing image synthesis framework that enables controllable generation guided by text descriptions and semantic masks. First, we propose a generalizable pipeline for constructing “image-text-segmentation” triplets by extracting semantic masks from OpenStreetMap (OSM) and generating text descriptions through large language models (LLMs), aiming to support the scalable construction of vision language datasets in remote sensing. Furthermore, building upon the ControlNet architecture, we effectively embed semantic masks into the encoder by multi-layer spatially-adaptive normalization operators, enabling the synthesized images to preserve geometric and spatial consistency with the semantic conditions. Experimental results demonstrate that our method achieves superior performance in generating high-quality remote sensing images both quantitatively and qualitatively compared to existing methods.
- 30 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
- Vol. 19 (19391404) , 16333-16345
Open AccessControllable Remote Sensing Image Generation With Semantic Consistency via Diffusion Models
The generation of remote sensing images plays a vital role in applications such as urban planning, urban development simulation, and environmental monitoring. The advent of diffusion models has introduced a new paradigm in image generation and achieved remarkable results across various fields. However, their adoption in remote sensing remains limited, primarily due to the lack of automated pipelines for constructing multimodal remote sensing datasets and the challenge of preserving semantic controllability given the complex object categories and high semantic similarity typical of remote sensing scenes. In this paper, we propose a remote sensing image synthesis framework that enables controllable generation guided by text descriptions and semantic masks. First, we propose a generalizable pipeline for constructing “image-text-segmentation” triplets by extracting semantic masks from OpenStreetMap (OSM) and generating text descriptions through large language models (LLMs), aiming to support the scalable construction of vision language datasets in remote sensing. Furthermore, building upon the ControlNet architecture, we effectively embed semantic masks into the encoder by multi-layer spatially-adaptive normalization operators, enabling the synthesized images to preserve geometric and spatial consistency with the semantic conditions. Experimental results demonstrate that our method achieves superior performance in generating high-quality remote sensing images both quantitatively and qualitatively compared to existing methods.
- 16 September 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Automation Science and Engineering
- Vol. 22 (15455955) , 6664-6676
Semi-Supervised Domain Adaptation With Dual-Adversarial Learning for Lane Detection
Lane detection approaches relying on abundant annotations have achieved great progress in autonomous driving. These approaches may suffer from performance collapse when directly applied to unseen images due to domain discrepancy. Most researches utilize unsupervised domain adaptation (UDA) to mitigate domain shifts. Despite this, there is still a large performance gap between UDA and supervised methods because lack of target supervision makes improving performance of cross-domain lane detection difficult. In this paper, we introduce semi-supervised domain adaptation (SSDA) to lane detection and develop a simple yet effective dual-adversarial learning scheme to minimize domain discrepancy. Specifically, we conduct global alignment by calculating entropy information of unlabeled target and source data to align entropy distribution and enhance feature confidence of lanes. Domain consistency adversarial learning is proposed to make full use of both source and unlabeled/labeled target data. A triplet of features is used to learn lane-invariant representations by distinguishing whether the feature belongs to same domain. This is conducive to promote consistency of lanes information in cross domain. In two datasets, our method obtains a better performance of 42.5% (54.9%) and 45.8% (64.1%) based on a small amount of target supervision. Experimental results demonstrate effectiveness of our method as well as its superiority to other methods in terms of reducing performance gap with supervised methods (https://github.com/shenhuqiji/LD-SSDA) Note to Practitioners —This paper focuses on the cross domain lane detection task to narrow domain discrepancy in lane detection among different scenarios for autonomous driving. It aims to reduce extra label annotations and retraining costs. Considering unsupervised domain adaptation method is hard to further improve detection performance when a large domain shift occurs in adaptation scenarios, we construct an effective dual-adversarial learning scheme based on semi-supervised domain adaptation for lane detection. This method takes as input a triplet of source data, unlabeled and labeled target data to effectively enhance cross domain detection accuracy, which is contributed to the perception tasks in driving scenarios, such as path and motion planning, lane departure warning, lane changing decisions. Our method has a better generalization ability in face of various scenarios and can be easily integrated in software platforms, which greatly facilitates the applications on intelligent systems in autonomous vehicles.
- 16 September 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Automation Science and Engineering
- Vol. 22 (15455955) , 6664-6676
Semi-Supervised Domain Adaptation With Dual-Adversarial Learning for Lane Detection
Lane detection approaches relying on abundant annotations have achieved great progress in autonomous driving. These approaches may suffer from performance collapse when directly applied to unseen images due to domain discrepancy. Most researches utilize unsupervised domain adaptation (UDA) to mitigate domain shifts. Despite this, there is still a large performance gap between UDA and supervised methods because lack of target supervision makes improving performance of cross-domain lane detection difficult. In this paper, we introduce semi-supervised domain adaptation (SSDA) to lane detection and develop a simple yet effective dual-adversarial learning scheme to minimize domain discrepancy. Specifically, we conduct global alignment by calculating entropy information of unlabeled target and source data to align entropy distribution and enhance feature confidence of lanes. Domain consistency adversarial learning is proposed to make full use of both source and unlabeled/labeled target data. A triplet of features is used to learn lane-invariant representations by distinguishing whether the feature belongs to same domain. This is conducive to promote consistency of lanes information in cross domain. In two datasets, our method obtains a better performance of 42.5% (54.9%) and 45.8% (64.1%) based on a small amount of target supervision. Experimental results demonstrate effectiveness of our method as well as its superiority to other methods in terms of reducing performance gap with supervised methods (https://github.com/shenhuqiji/LD-SSDA) Note to Practitioners —This paper focuses on the cross domain lane detection task to narrow domain discrepancy in lane detection among different scenarios for autonomous driving. It aims to reduce extra label annotations and retraining costs. Considering unsupervised domain adaptation method is hard to further improve detection performance when a large domain shift occurs in adaptation scenarios, we construct an effective dual-adversarial learning scheme based on semi-supervised domain adaptation for lane detection. This method takes as input a triplet of source data, unlabeled and labeled target data to effectively enhance cross domain detection accuracy, which is contributed to the perception tasks in driving scenarios, such as path and motion planning, lane departure warning, lane changing decisions. Our method has a better generalization ability in face of various scenarios and can be easily integrated in software platforms, which greatly facilitates the applications on intelligent systems in autonomous vehicles.
- Open Access
- 12 May 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Vehicular Technology
- Vol. 72 (10) , 12602-12614
MLP-Based Efficient Convolutional Neural Network for Lane Detection
Lane detection is an important and fundamental task in autonomous driving. Modern convolutional neural network (CNN) methods have achieved high performance in lane detection; however, the intrinsic locality of convolution operations makes these methods limited in effectively modeling the long-range dependencies that are vital to capture global information of lanes. Additionally, numerous convolution operations result in considerable computational cost for high complexity. To overcome these difficulties, we propose an efficient lane detection method by combining CNN with a multilayer perceptron (MLP). First, an improved bottleneck-1D layer is used to replace the standard convolutional layer in overall network to reduce the computational cost and parameters while applying hybrid dilated convolution (HDC) to better capture multiscale lane information. Second, we construct a hybrid MLP block in the latent space to capture the long-range dependencies of lanes. The hybrid MLP projects tokenized convolutional features from spatial locations and channels, and then, they are fused together to obtain global representation, in which each output pixel is related to each input pixel. The introduction of MLP further decreases computational complexity and makes the proposed architecture more efficient for lane detection. Experimental results on two challenging datasets (CULane, Tusimple) demonstrate that our method can achieve a higher computational efficiency while maintaining a decent detection performance compared with other state-of-the-art methods. Furthermore, this study indicates that integrating the global representation capacity of an MLP with local prior information of convolution is an effective and potential perspective in lane detection.
- 12 May 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Vehicular Technology
- Vol. 72 (10) , 12602-12614
Open AccessMLP-Based Efficient Convolutional Neural Network for Lane Detection
Lane detection is an important and fundamental task in autonomous driving. Modern convolutional neural network (CNN) methods have achieved high performance in lane detection; however, the intrinsic locality of convolution operations makes these methods limited in effectively modeling the long-range dependencies that are vital to capture global information of lanes. Additionally, numerous convolution operations result in considerable computational cost for high complexity. To overcome these difficulties, we propose an efficient lane detection method by combining CNN with a multilayer perceptron (MLP). First, an improved bottleneck-1D layer is used to replace the standard convolutional layer in overall network to reduce the computational cost and parameters while applying hybrid dilated convolution (HDC) to better capture multiscale lane information. Second, we construct a hybrid MLP block in the latent space to capture the long-range dependencies of lanes. The hybrid MLP projects tokenized convolutional features from spatial locations and channels, and then, they are fused together to obtain global representation, in which each output pixel is related to each input pixel. The introduction of MLP further decreases computational complexity and makes the proposed architecture more efficient for lane detection. Experimental results on two challenging datasets (CULane, Tusimple) demonstrate that our method can achieve a higher computational efficiency while maintaining a decent detection performance compared with other state-of-the-art methods. Furthermore, this study indicates that integrating the global representation capacity of an MLP with local prior information of convolution is an effective and potential perspective in lane detection.
- Open Access
- 31 March 2023
- Published by MDPI AG in Remote Sensing
- Vol. 15 (7) , 1887
Estimation of Daily Arctic Winter Sea Ice Thickness from Thermodynamic Parameters Using a Self-Attention Convolutional Neural Network
Thermodynamic parameters play a crucial role in determining polar sea ice thickness (SIT); however, modeling their relationship is difficult due to the complexity of the influencing mechanisms. In this study, we propose a self-attention convolutional neural network (SAC-Net), which aims to model the relationship between thermodynamic parameters and SIT more parsimoniously, allowing us to estimate SIT directly from these parameters. SAC-Net uses a fully convolutional network as a baseline model to detect the spatial information of the thermodynamic parameters. Furthermore, a self-attention block is introduced to enhance the correlation among features. SAC-Net was trained on a dataset of SIT observations and thermodynamic data from the 2012–2019 freeze-up period, including surface upward sensible heat flux, surface upward latent heat flux, 2 m temperature, skin temperature, and surface snow temperature. The results show that our neural network model outperforms two thermodynamic-based SIT products in terms of accuracy and can provide reliable estimates of SIT. This study demonstrates the potential of the neural network to provide accurate and automated predictions of Arctic winter SIT from thermodynamic data, and, thus, the network can be used to support decision-making in certain fields, such as polar shipping, environmental protection, and climate science.
- 31 March 2023
- Published by MDPI AG in Remote Sensing
- Vol. 15 (7) , 1887
Open AccessEstimation of Daily Arctic Winter Sea Ice Thickness from Thermodynamic Parameters Using a Self-Attention Convolutional Neural Network
Thermodynamic parameters play a crucial role in determining polar sea ice thickness (SIT); however, modeling their relationship is difficult due to the complexity of the influencing mechanisms. In this study, we propose a self-attention convolutional neural network (SAC-Net), which aims to model the relationship between thermodynamic parameters and SIT more parsimoniously, allowing us to estimate SIT directly from these parameters. SAC-Net uses a fully convolutional network as a baseline model to detect the spatial information of the thermodynamic parameters. Furthermore, a self-attention block is introduced to enhance the correlation among features. SAC-Net was trained on a dataset of SIT observations and thermodynamic data from the 2012–2019 freeze-up period, including surface upward sensible heat flux, surface upward latent heat flux, 2 m temperature, skin temperature, and surface snow temperature. The results show that our neural network model outperforms two thermodynamic-based SIT products in terms of accuracy and can provide reliable estimates of SIT. This study demonstrates the potential of the neural network to provide accurate and automated predictions of Arctic winter SIT from thermodynamic data, and, thus, the network can be used to support decision-making in certain fields, such as polar shipping, environmental protection, and climate science.
- Open Access
- 18 August 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
- Vol. 14 (19391404) , 8407-8418
Weakly-Supervised Domain Adaptation With Adversarial Entropy for Building Segmentation in Cross-Domain Aerial Imagery
Building segmentation is a classical and challenging task in high-resolution remote sensing imagery. This approach has achieved remarkable performance based on a fully convolutional network with adequate pixel-wise annotations. However, due to differences in sensor technology as well as appearance in different regions, datasets gathered from these various sources are quite distinct, and dense annotations for a particular area are not always available. Thus, directly applying a segmentation model trained on one dataset (source domain) to another unseen dataset (target domain) usually results in a drop in performance, called the domain gap. In this article, we propose a weakly-supervised domain adaptation method using adversarial entropy for building segmentation to address this problem. First, we use an adversarial entropy strategy to decrease the entropy and improve the prediction certainty for target images, causing the distributions between the source and target domains to become closer to each other. Second, we propose a simple and effective self-training strategy for the target domain that produces high-confidence predictions using pseudolabels. We use a series of thresholds to generate the pseudolabels without introducing extra parameters. This strategy effectively enhances the discriminability of the target domain and further minimizes the distribution discrepancy between the two domains. Experiments on cross-domain aerial datasets have demonstrated the effectiveness and superiority of our proposed method when compared to other state-of-the-art methods.
- 18 August 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
- Vol. 14 (19391404) , 8407-8418
Open AccessWeakly-Supervised Domain Adaptation With Adversarial Entropy for Building Segmentation in Cross-Domain Aerial Imagery
Building segmentation is a classical and challenging task in high-resolution remote sensing imagery. This approach has achieved remarkable performance based on a fully convolutional network with adequate pixel-wise annotations. However, due to differences in sensor technology as well as appearance in different regions, datasets gathered from these various sources are quite distinct, and dense annotations for a particular area are not always available. Thus, directly applying a segmentation model trained on one dataset (source domain) to another unseen dataset (target domain) usually results in a drop in performance, called the domain gap. In this article, we propose a weakly-supervised domain adaptation method using adversarial entropy for building segmentation to address this problem. First, we use an adversarial entropy strategy to decrease the entropy and improve the prediction certainty for target images, causing the distributions between the source and target domains to become closer to each other. Second, we propose a simple and effective self-training strategy for the target domain that produces high-confidence predictions using pseudolabels. We use a series of thresholds to generate the pseudolabels without introducing extra parameters. This strategy effectively enhances the discriminability of the target domain and further minimizes the distribution discrepancy between the two domains. Experiments on cross-domain aerial datasets have demonstrated the effectiveness and superiority of our proposed method when compared to other state-of-the-art methods.