TOHOKU UNIVERSITY Researchers

Details of the Researcher

Home

日本語 English

Wei Yuan

Section

International Research Institute of Disaster Science

Job title

Associate Professor

Degree

Ph.D (The University of Tokyo)
Master of Engineering (The University of Tokyo)
Doctor of Engineering (Wuhan University)

researchmap

https://researchmap.jp/milowei304

J-GLOBAL ID

202101007307066759

e-Rad No.

60837475

ORCID

https://orcid.org/0000-0002-1370-0079

Research History 4

2026/07 - Present

Tohoku University　Frontier Research Institute for Interdisciplinary Sciences　Fellow
2024/02 - Present

Tohoku University
2020/03 - 2024/02

The University of Tokyo
2018/10 - 2020/03

東京大学

Education 3

Wuhan University　School of Remote Sensing and Information Engineering　Photogrammetry and Remote Sensing

2012/09 - 2020/06
The University of Tokyo　The Graduate School of Engineering　Department of Civil Engineering

2015/10 - 2018/09
The University of Tokyo　The Graduate School of Engineering　Department of Civil Engineering

2014/10 - 2015/09

Professional Memberships 4

IEEE

2018/07 - Present
IEEE Geoscience and Remote Sensing Society

2018/06 - Present
American Society for Photogrammetry and Remote Sensing

2018/06 - Present
International Society of Photogrammetry and Remote Sensing

2016/06 - Present

Research Interests 1

Photogrammetry;Remote Sensing;3D Reconstruction;Disaster Assesment;GeoAI

Research Areas 1

Social infrastructure (civil Engineering, architecture, disaster prevention) / Civil engineering (planning and transportation) /

Awards 3

ISPRS WEC Kennert Torlegård Award

2026/07　International Society of Photogrammetry and Remote Sensing　Flood Depth Mapping from SAR Imagery Using CS-Mamba with DEM Sensitivity Analysis
Best Young Researcher Award

2024/10　International Society of Photogrammetry and Remote Sensing Commission IV　The Performance of the Optical Flow Field based Dense Image Matching for UAV Imagery
Best Young Author paper award

2022/06　International Society for Photogrammetry and Remote Sensing　Learning Social Compliant Multi-Modal Distributions of Human Path in Crowds

Papers 48

Kinematic Characteristics and Risk Analysis of Potential Rockfall based on 3D Point Clouds Peer-reviewed

Wei Yuan, Changqing Liu, Han Bao, Weihang Ran, Zhongyuan Yang, Xiuxiao Yuan, Ryosuke Shibasaki, Shunichi Koshimura

The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences　XLIX-B4-2026　223-228　2026/08/04
Publisher: Copernicus GmbH
DOI： 10.5194/isprs-archives-xlix-b4-2026-223-2026 　

eISSN： 2194-9034

More details Close

Abstract. Rockfall hazard on steep rock slopes is strongly controlled by the geometry of in-situ blocks bounded by intersecting discontinuities. In practice, however, source blocks are still commonly identified by field observation and manual interpretation, which is difficult to apply consistently on high, steep slopes. This study develops a three-dimensional point-cloud-based workflow for locating, characterising, and assessing potentially unstable rock blocks under real terrain conditions. The approach combines structural interpretation with visual kinematic criteria to detect discontinuity combinations associated with planar sliding, wedge sliding, and toppling, and to map unstable blocks directly in point-cloud space. Block volumes are then estimated by point-cloud differencing, allowing the number and size distribution of source blocks to be described quantitatively. Finally, representative blocks are introduced into three-dimensional rockfall simulations using their measured location and volume. The simulations provide rockfall frequency, bounce height, velocity, and kinetic energy, and are used to evaluate the threat posed to the transportation corridor below the slope. The proposed workflow links source identification with dynamic hazard assessment and offers a practical basis for rapid screening and refined mitigation of rockfall-prone slopes in mountainous infrastructure corridors.
Flood Depth Mapping from SAR Imagery Using CS-Mamba with DEM Sensitivity Analysis Peer-reviewed

Zhongyuan Yang, Wei Yuan, Weihang Ran, Changqing Liu, Bruno Adriano, Ryosuke Shibasaki, Shunichi Koshimura

ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences　XI-3-2026　493-500　2026/07/08
Publisher: Copernicus GmbH
DOI： 10.5194/isprs-annals-xi-3-2026-493-2026 　

eISSN： 2194-9050

More details Close

Abstract. Accurate flood extent and depth information are essential to emergency response, yet most existing studies treat these tasks separately. This work introduces an integrated SAR-to-depth framework that combines water body semantic segmentation with DEM-based geometric depth estimation to generate both flood-extent maps and pixel-wise depth products from Sentinel-1 imagery. For flood extent mapping, we propose a cross-scale Mamba with selective state-space blocks, which achieves a mean IoU of 79.8% across ten European flood events from the KuroSiwo benchmark, outperforming RSMamba by 7.4% and surpassing common CNN baselines. The experimental results demonstrate that the proposed model also generalizes well to unseen events, with test performance exceeding validation scores. When both CS-Mamba predictions and KuroSiwo reference masks are input to FLEXTH, the resulting depth estimates agree within ±2% across four global DEMs. Initial validation against ICESat-2 altimetry using MERIT DEM (19 matched points) shows RMSE of 4.60 m and Bias of -1.88 m, providing preliminary validation evidence with systematic underestimation. Systematic DEM comparison shows FLEXTH is robust across all four DEMs, with Copernicus and MERIT showing closest agreement with reference mask estimates. The framework produces three-class flood masks and pixel-wise depth maps, combining extent mapping with quantitative depth information for operational flood monitoring.
An efficient cost calculation method for disparity estimation of stereo matching considering shadow occlusion Peer-reviewed

Hongjun Sha, Wei Yuan, Xunping Wang, Xiuliu Yuan, Shunichi Koshimura

Geo-spatial Information Science　2026/01/02

DOI： 10.1080/10095020.2025.2487138 　
Taming Spatial Heterophily and Temporal Irregularity: A Curriculum Learning Approach for Traffic Forecasting Peer-reviewed

Hongjun Wang, Zhiwen Zhang, Jiyuan Chen, Zipei Fan, Renhe Jiang, Wei Yuan, Ryosuke Shibasaki, Xuan Song

IEEE Transactions on Intelligent Transportation Systems　1-17　2026
Publisher: Institute of Electrical and Electronics Engineers (IEEE)
DOI： 10.1109/tits.2026.3685281 　

ISSN： 1524-9050

eISSN： 1558-0016
Visibility-Aware Disparity Estimation for Aerial Images by Fusing Line Features in Shadow Areas Peer-reviewed

Hongjun Sha, Wei Yuan, Siyuan Zou, Ryosuke Shibasaki, Shunichi Koshimura

IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing　1-26　2026
Publisher: Institute of Electrical and Electronics Engineers (IEEE)
DOI： 10.1109/jstars.2026.3683686 　

ISSN： 1939-1404

eISSN： 2151-1535
Quantitative identification and hazard assessment of potentially unstable rock masses along mountainous transportation corridors Peer-reviewed

Changqing Liu, Han Bao, Jingfeng Zhang, Jinghao Yang, Xuanyan Dong, Hengxing Lan, Shunichi Koshimura, Wei Yuan

Engineering Failure Analysis　2026/01

DOI： 10.1016/j.engfailanal.2025.110273 　
Intelligent characterization of discontinuities and heterogeneity evaluation of potential hazard sources in high-steep rock slope by TLS-UAV technology Peer-reviewed

Changqing Liu, Han Bao, Tianyi Wang, Jingfeng Zhang, Hengxing Lan, Shengwen Qi, Wei Yuan, Shunichi Koshimura

Journal of Rock Mechanics and Geotechnical Engineering　2026/01

DOI： 10.1016/j.jrmge.2025.03.023 　
PortVIS: An Interactive Platform for Port-to-Port Trajectory Imputation and Visual Analytics Peer-reviewed

Zhiwen Zhang, Zipei Fan, Wei Yuan, Shun Iwazaki, Ryosuke Shibasaki

Proceedings of the 33rd ACM International Conference on Advances in Geographic Information Systems　812-815　2025/11/03
Publisher: ACM
DOI： 10.1145/3748636.3762787 　
Accurate Digital Reconstruction of High-Steep Rock Slope via Transformer-Based Multi-Sensor Data Fusion Peer-reviewed

Changqing Liu, Han Bao, Jingfeng Zhang, Hengxing Lan, Bruno Adriano, Shunichi Koshimura, Wei Yuan

Remote Sensing　17　(21)　3555-3555　2025/10/28
Publisher: MDPI AG
DOI： 10.3390/rs17213555 　

eISSN： 2072-4292

More details Close

Accurate and comprehensive characterization of high-steep slopes is crucial for real-time risk prediction, disaster assessment, and damage evolution monitoring. The study focused on a high-steep rocky slope along the Yanjiang Expressway in Sichuan Province, China. A novel digital reconstruction method was introduced, which integrates terrestrial laser scanning (TLS) and unmanned aerial vehicle (UAV) photogrammetry through a Transformer-based method combining GeoTransformer with the Maximal Cliques (MAC) algorithm. The results indicated that TLS excels in capturing fine-scale features, whereas UAV demonstrates superior performance in large-scale terrain reconstruction. However, multi-sensor data exhibit heterogeneity in terms of partial overlap, large outliers, and density differences. To address these challenges, the GeoTransformer-MAC framework extracts geometrically invariant features from cross-source point cloud (CSPC) to establish initial correspondences, followed by rigorous screening of high-quality locally consistent correspondences to optimize transformation parameters. This method achieves accurate digital reconstruction of the high-steep rock slope. Global and local error analyses verify the model’s superiority in both overall slope characterization and fine-scale feature representation. Compared with the TLS-only model and the conventional method, the Transformer-based method improves the slope model integrity by 85.58%, increases the data density by 9.71%, and improves the accuracy by nearly threefold. This study provides a novel approach for the digital modeling of complex terrains, which serves the refined identification and modeling of geohazards for high-steep slopes in complex mountainous regions.
A Generalized Deep Learning Method for Rooftop Condition Assessment via Monocular Imagery Peer-reviewed

Xinyu Li, Zhiling Guo, Jian Xu, Wei Yuan, Xiaoya Song, Haoran Zhang, Jinyue Yan

IGARSS 2025 - 2025 IEEE International Geoscience and Remote Sensing Symposium　790-793　2025/08/03
Publisher: IEEE
DOI： 10.1109/igarss55030.2025.11243709 　
Multi-source 3D point clouds fusion for potential rock mass hazard evaluation in high-steep rock slopes Peer-reviewed

Wei Yuan, Changqing Liu, Tianyi Wang, Bruno Adriano, Han Bao, Ryosuke Shibasaki, Shunichi Koshimura

The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences　XLVIII-G-2025　1663-1668　2025/08/02
Publisher: Copernicus GmbH
DOI： 10.5194/isprs-archives-xlviii-g-2025-1663-2025 　

eISSN： 2194-9034

More details Close

Abstract. Accurate characterization and evaluation of hazardous rockmass sources prove essential for rockfall risk mitigation. Structural properties of rock masses play a decisive role in evaluating these risks. This study presents an integrated approach that combines Terrestrial Laser Scanning (TLS) and Unmanned Aerial Vehicle (UAV) photogrammetry to address data limitations in complex terrain. The practical validation was carried out on the basis of a case study on a high and steep rock slope. The results demonstrate that the fusion of TLS-UAV multi-source data enhances spatial coverage and point cloud density by 19%, enabling comprehensive slope modeling and improving multi-angle structural characterization of target rock masses. An approach integrating multiple algorithms enables the automatic identification of rock joints from multi-source 3D point clouds, achieving high recognition accuracy. And the key geometric and mechanical parameters were extracted and analyzed to quantify joint properties. Furthermore, a novel rock hazard index (RHI) is proposed, which takes into account joint geometric features, joint mechanical features, and slope quality grade to assess risk levels across slope domains. The proposed framework provides an efficient solution for joint-controlled hazardous rockmass assessment, offering theoretical insights and practical applications for infrastructure-related geohazard prevention. This study contributes to enhancing risk assessment methodologies for high and steep slope environments.
Tiered Spatio-Temporal Difficulty: Curriculum Scheduler for Multi-Sensor Traffic Flow Prediction Peer-reviewed

Zhiwen Zhang, Hongjun Wang, Zipei Fan, Renhe Jiang, Wei Yuan, Xuan Song, Ryosuke Shibasaki

IEEE Transactions on Mobile Computing　1-15　2025
Publisher: Institute of Electrical and Electronics Engineers (IEEE)
DOI： 10.1109/tmc.2025.3608620 　

ISSN： 1536-1233

eISSN： 1558-0660 2161-9875
AISFuser: Encoding Maritime Graphical Representations With Temporal Attribute Modeling for Vessel Trajectory Prediction Peer-reviewed

Zhiwen Zhang, Wei Yuan, Zipei Fan, Xuan Song, Ryosuke Shibasaki

IEEE Transactions on Knowledge and Data Engineering　2025

DOI： 10.1109/TKDE.2025.3531770 　
The Performance of the Optical Flow Field based Dense Image Matching for UAV Imagery Peer-reviewed

Wei Yuan, Weihang Ran, Bruno Adriano, Ryosuke Shibasaki, Shunichi Koshimura

ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences　X-4-2024　433-440　2024/10/18
Publisher: Copernicus GmbH
DOI： 10.5194/isprs-annals-x-4-2024-433-2024 　

eISSN： 2194-9050

More details Close

Abstract. With the rapid development of sensing platforms, unmanned aerial vehicle (UAV)-based mapping has become increasingly popular because of its economic efficiency and flexibility, especially for providing 3D information to support urban growth monitoring and change detection to meet sustainable development goals (SDGs). This paper presents an improved optical flow field-based dense matching algorithm (OFFDM) for low-altitude UAV images based on the Ph.D. thesis of Yuan (Yuan, 2018). First, high-precision seed points were used to compute the optical flow field within stereo pairs, effectively minimizing redundant calculations during the fine-matching phase. Second, a fine-matching approach, integrating multiple constraints, was applied to refine the coarse matching results based on the optical flow field. Extensive dense matching experiments on UAV low-altitude aerial imagery assessed the performance of OFFDIM across four dimensions: 3D point cloud visualization, matching success rate, precision, and reliability. Extensive experiments on low-altitude UAV imagery, characterized by a resolution of 7cm per pixel over a 10,608×8,608 pixel dimension and a 60% forward overlap, evaluate the OFFDM's efficacy. The quantitative evaluation revealed that the proposed method achieved an accuracy of ±0.7 pixels in image coordinates and ±20 cm on the ground, with a matching success rate exceeding 97%. The processing time was approximately 272 seconds for handling one single stereo pair. When compared to the widely adopted PMVS algorithm, known for its effectiveness in dense matching for UAV images, the proposed method demonstrated higher completeness and improved matching efficiency by more than five times. These results demonstrated that the proposed approach is more suitable for dense matching on UAV imagery-based high-precision 3D spatial data extraction, supporting global mapping tasks more effectively.
General generative AI-based image augmentation method for robust rooftop PV segmentation Peer-reviewed

Hongjun Tan, Zhiling Guo, Zhengyuan Lin, Yuntian Chen, Dou Huang, Wei Yuan, Haoran Zhang, Jinyue Yan

Applied Energy　368　123554-123554　2024/08
Publisher: Elsevier BV
DOI： 10.1016/j.apenergy.2024.123554 　

ISSN： 0306-2619
Nuclei-level prior knowledge constrained multiple instance learning for breast histopathology whole slide image classification Peer-reviewed

Xunping Wang, Wei Yuan

iScience　27　(6)　109826-109826　2024/06
Publisher: Elsevier BV
DOI： 10.1016/j.isci.2024.109826 　

ISSN： 2589-0042
Hybrid Network-Based Automatic Seamline Detection for Orthophoto Mosaicking Peer-reviewed

Wei Yuan, Yang Cai, Jonathan Li

IEEE Transactions on Geoscience and Remote Sensing　62　1-14　2024/04
Publisher: Institute of Electrical and Electronics Engineers (IEEE)
DOI： 10.1109/tgrs.2024.3393626 　

ISSN： 0196-2892

eISSN： 1558-0644
A cluster-based disambiguation method using pose consistency verification for structure from motion Peer-reviewed

Ye Gong, Pengwei Zhou, Changfeng Liu, Yan Yu, Jian Yao, Wei Yuan, Li Li

ISPRS Journal of Photogrammetry and Remote Sensing　2024/03

DOI： 10.1016/j.isprsjprs.2024.02.016 　
Hybrid Feature Embedding for Automatic Building Outline Extraction Peer-reviewed

Weihang Ran, Wei Yuan, Xiaodan Shi, Zipei Fan, Ryosuke Shibasaki

IGARSS 2023 - 2023 IEEE International Geoscience and Remote Sensing Symposium　2023/07/16
Publisher: IEEE
DOI： 10.1109/igarss52108.2023.10282867 　
Graph Encoding based Hybrid Vision Transformer for Automatic Road Network Extraction Peer-reviewed

Wei Yuan, Weihang Ran, Xiaodan Shi, Zipei Fan, Yang Cai, Ryosuke Shibasaki

IGARSS 2023 - 2023 IEEE International Geoscience and Remote Sensing Symposium　2023/07/16
Publisher: IEEE
DOI： 10.1109/igarss52108.2023.10283247 　
Few-Shot Depth Completion Using Denoising Diffusion Probabilistic Model Peer-reviewed

Weihang Ran, Wei Yuan, Ryosuke Shibasaki

2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW)　2023-June　6559-6567　2023/06
Publisher: IEEE
DOI： 10.1109/cvprw59228.2023.00697 　

ISSN： 2160-7508

eISSN： 2160-7516
LiteST-Net: A Hybrid Model of Lite Swin Transformer and Convolution for Building Extraction from Remote Sensing Image Peer-reviewed

Wei Yuan, Xiaobo Zhang, Jibao Shi, Jin Wang

Remote Sensing　15　(8)　1996-1996　2023/04/10
Publisher: MDPI AG
DOI： 10.3390/rs15081996 　

eISSN： 2072-4292

More details Close

Extracting building data from remote sensing images is an efficient way to obtain geographic information data, especially following the emergence of deep learning technology, which results in the automatic extraction of building data from remote sensing images becoming increasingly accurate. A CNN (convolution neural network) is a successful structure after a fully connected network. It has the characteristics of saving computation and translation invariance with improved local features, but it has difficulty obtaining global features. Transformers can compensate for the shortcomings of CNNs and more effectively obtain global features. However, the calculation number of transformers is excessive. To solve this problem, a Lite Swin transformer is proposed. The three matrices Q, K, and V of the transformer are simplified to only a V matrix, and the v of the pixel is then replaced by the v with the largest projection value on the pixel feature vector. In order to better integrate global features and local features, we propose the LiteST-Net model, in which the features extracted by the Lite Swin transformer and the CNN are added together and then sampled up step by step to fully utilize the global feature acquisition ability of the transformer and the local feature acquisition ability of the CNN. The comparison experiments on two open datasets are carried out using our proposed LiteST-Net and some classical image segmentation models. The results show that compared with other networks, all metrics of LiteST-Net are the best, and the predicted image is closer to the label.
Voronoi Centerline-Based Seamline Network Generation Method Peer-reviewed

Xiuxiao Yuan, Yang Cai, Wei Yuan

Remote Sensing　15　(4)　2023/02

DOI： 10.3390/rs15040917 　

eISSN： 2072-4292
Multiconstraint Transformer-Based Automatic Building Extraction From High-Resolution Remote Sensing Images Peer-reviewed

Wei Yuan, Weihang Ran, Xiaodan Shi, Ryosuke Shibasaki

IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing　2023

DOI： 10.1109/JSTARS.2023.3319826 　
MetaTraj: Meta-Learning for Cross-Scene Cross-Object Trajectory Prediction Peer-reviewed

Xiaodan Shi, Haoran Zhang, Wei Yuan, Ryosuke Shibasaki

IEEE Transactions on Intelligent Transportation Systems　1-10　2023
Publisher: Institute of Electrical and Electronics Engineers ({IEEE})
DOI： 10.1109/TITS.2023.3299112 　

ISSN： 1524-9050

eISSN： 1558-0016
Fully automatic DOM generation method based on optical flow field dense image matching Peer-reviewed

Wei Yuan, Xiuxiao Yuan, Yang Cai, Ryosuke Shibasaki

Geo-Spatial Information Science　26　(2)　242-256　2023

DOI： 10.1080/10095020.2022.2159886 　

ISSN： 1009-5020
Shift Pooling PSPNet: Rethinking PSPNet for Building Extraction in Remote Sensing Images from Entire Local Feature Pooling Peer-reviewed

Wei Yuan, Jin Wang, Wenbo Xu

Remote Sensing　14　(19)　4889-4889　2022/09/30
Publisher: MDPI AG
DOI： 10.3390/rs14194889 　

eISSN： 2072-4292

More details Close

Building extraction by deep learning from remote sensing images is currently a research hotspot. PSPNet is one of the classic semantic segmentation models and is currently adopted by many applications. Moreover, PSPNet can use not only CNN-based networks but also transformer-based networks as backbones; therefore, PSPNet also has high value in the transformer era. The core of PSPNet is the pyramid pooling module, which gives PSPNet the ability to capture the local features of different scales. However, the pyramid pooling module also has obvious shortcomings. The grid is fixed, and the pixels close to the edge of the grid cannot obtain the entire local features. To address this issue, an improved PSPNet network architecture named shift pooling PSPNet is proposed, which uses a module called shift pyramid pooling to replace the original pyramid pooling module, so that the pixels at the edge of the grid can also obtain the entire local features. Shift pooling is not only useful for PSPNet but also in any network that uses a fixed grid for downsampling to increase the receptive field and save computing, such as ResNet. A dense connection was adopted in decoding, and upsampling was gradually carried out. With two open datasets, the improved PSPNet, PSPNet, and some classic image segmentation models were used for comparative experiments. The results show that our method is the best according to the evaluation metrics, and the predicted image is closer to the label.
GapLoss: A Loss Function for Semantic Segmentation of Roads in Remote Sensing Images Peer-reviewed

Wei Yuan, Wenbo Xu

Remote Sensing　14　(10)　2422-2422　2022/05/18
Publisher: MDPI AG
DOI： 10.3390/rs14102422 　

eISSN： 2072-4292

More details Close

At present, road continuity is a major challenge, and it is difficult to extract the centerline vector of roads, especially when the road view is obstructed by trees or other structures. Most of the existing research has focused on optimizing the available deep-learning networks. However, the segmentation accuracy is also affected by the loss function. Currently, little research has been published on road segmentation loss functions. To resolve this problem, an attention loss function named GapLoss that can be combined with any segmentation network was proposed. Firstly, a deep-learning network was used to obtain a binary prediction mask. Secondly, a vector skeleton was extracted from the prediction mask. Thirdly, for each pixel, eight neighboring pixels with the same value of the pixel were calculated. If the value was 1, then the pixel was identified as the endpoint. Fourth, according to the number of endpoints within a buffered range, each pixel in the prediction image was given a corresponding weight. Finally, the weighted average value of the cross-entropy of all the pixels in the batch was used as the final loss function value. We employed four well-known semantic segmentation networks to conduct comparative experiments on three large datasets. The results showed that, compared to other loss functions, the evaluation metrics after using GapLoss were nearly all improved. From the predicted image, the road prediction by GapLoss was more continuous, especially at intersections and when the road was obscured from view, and the road segmentation accuracy was improved.
LEARNING SOCIAL COMPLIANT MULTI-MODAL DISTRIBUTIONS OF HUMAN PATH IN CROWDS Peer-reviewed

X. Shi, H. Zhang, W. Yuan, D. Huang, Z. Guo, R. Shibasaki

ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences　V-4-2022　(4)　91-98　2022/05/18
Publisher: Copernicus {GmbH}
DOI： 10.5194/isprs-annals-v-4-2022-91-2022 　

ISSN： 2194-9050

eISSN： 2194-9050
Online trajectory prediction for metropolitan scale mobility digital twin. Peer-reviewed

Zipei Fan, Xiaojie Yang, Wei Yuan, Renhe Jiang, Quanjun Chen, Xuan Song 0001, Ryosuke Shibasaki

SIGSPATIAL/GIS　103-12　2022

DOI： 10.1145/3557915.3561040 　
Impact of sensor data sampling rate in gnss/ins integrated navigation with various sensor grades Peer-reviewed

Y. Wei, Y. Li

International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences - ISPRS Archives　46　(3/W1-2022)　205-211　2022

DOI： 10.5194/isprs-archives-XLVI-3-W1-2022-205-2022 　

ISSN： 1682-1750
Cross-Scale Attention-based Tree Crown Detection via UAV imagery Peer-reviewed

Wei Yuan, Xiaodan Shi, Zhiling Guo, Zipei Fan, Jianya Gong, Ryosuke Shibasaki

International Geoscience and Remote Sensing Symposium (IGARSS)　2022-July　2203-2206　2022

DOI： 10.1109/IGARSS46834.2022.9884316 　
MSST-Net: A Multi-Scale Adaptive Network for Building Extraction from Remote Sensing Images Based on Swin Transformer Peer-reviewed

Wei Yuan, Wenbo Xu

Remote Sensing　13　(23)　4743-4743　2021/11/23
Publisher: MDPI AG
DOI： 10.3390/rs13234743 　

eISSN： 2072-4292

More details Close

The segmentation of remote sensing images by deep learning technology is the main method for remote sensing image interpretation. However, the segmentation model based on a convolutional neural network cannot capture the global features very well. A transformer, whose self-attention mechanism can supply each pixel with a global feature, makes up for the deficiency of the convolutional neural network. Therefore, a multi-scale adaptive segmentation network model (MSST-Net) based on a Swin Transformer is proposed in this paper. Firstly, a Swin Transformer is used as the backbone to encode the input image. Then, the feature maps of different levels are decoded separately. Thirdly, the convolution is used for fusion, so that the network can automatically learn the weight of the decoding results of each level. Finally, we adjust the channels to obtain the final prediction map by using the convolution with a kernel of 1 × 1. By comparing this with other segmentation network models on a WHU building data set, the evaluation metrics, mIoU, F1-score and accuracy are all improved. The network model proposed in this paper is a multi-scale adaptive network model that pays more attention to the global features for remote sensing segmentation.
GRAPH NEURAL NETWORK BASED MULTI-FEATURE FUSION FOR BUILDING CHANGE DETECTION Peer-reviewed

W. Yuan, X. Yuan, Z. Fan, Z. Guo, X. Shi, J. Gong, R. Shibasaki

The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences　{XLIII}-B3-2021　(B3-2021)　377-382　2021/06/28
Publisher: Copernicus {GmbH}
DOI： 10.5194/isprs-archives-xliii-b3-2021-377-2021 　

ISSN： 2194-9034
END-TO-END BUILDING CHANGE DETECTION MODEL IN AERIAL IMAGERY AND DIGITAL SURFACE MODEL BASED ON NEURAL NETWORKS Peer-reviewed

X. Lian, W. Yuan, Z. Guo, Z. Cai, X. Song, R. Shibasaki

ISPRS - International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences　{XLIII}-B2-2020　(B2)　1239-1246　2020/08/14
Publisher: Copernicus {GmbH}
DOI： 10.5194/isprs-archives-xliii-b2-2020-1239-2020 　

ISSN： 2194-9034
UNSUPERVISED MULTI-CONSTRAINT DEEP NEURAL NETWORK FOR DENSE IMAGE MATCHING Peer-reviewed

W. Yuan, Z. Fan, X. Yuan, J. Gong, R. Shibasaki

ISPRS - International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences　{XLIII}-B2-2020　(B2)　163-167　2020/08/12
Publisher: Copernicus {GmbH}
DOI： 10.5194/isprs-archives-xliii-b2-2020-163-2020 　

ISSN： 2194-9034
Multimodal Interaction-Aware Trajectory Prediction in Crowded Space Peer-reviewed

Xiaodan Shi, Xiaowei Shao, Zipei Fan, Renhe Jiang, Haoran Zhang, Zhiling Guo, Guangming Wu, Wei Yuan, Ryosuke Shibasaki

Proceedings of the AAAI Conference on Artificial Intelligence　34　(07)　11982-11989　2020/04/03
Publisher: Association for the Advancement of Artificial Intelligence ({AAAI})
DOI： 10.1609/aaai.v34i07.6874 　

ISSN： 2159-5399
Super-resolution integrated building semantic segmentation for multi-source remote sensing imagery Peer-reviewed

Guo, Z., Wu, G., Song, X., Yuan, W., Chen, Q., Zhang, H., Shi, X., Xu, M., Xu, Y., Shibasaki, R., Shao, X.

IEEE Access　7　99381-99397　2019

DOI： 10.1109/ACCESS.2019.2928646 　

ISSN： 2169-3536
Research developments and prospects on dense image matching in photogrammetry,航摄影像密集匹配的研究进展与展望 Peer-reviewed

Yuan, X., Yuan, W., Xu, S., Ji, Y.

Cehui Xuebao/Acta Geodaetica et Cartographica Sinica　48　(12)　1542-1550　2019

DOI： 10.11947/j.AGCS.2019.20190453 　

ISSN： 1001-1595
Dense image-matching via optical flow field estimation and fast-guided filter refinement Peer-reviewed

Yuan, W., Yuan, X., Xu, S., Gong, J., Shibasaki, R.

Remote Sensing　11　(20)　2410-2410　2019

DOI： 10.3390/rs11202410 　

ISSN： 2072-4292
Semantic segmentation for urban planning maps based on U-Net Peer-reviewed

Zhiling Guo, Hiroaki Shengoku, Guangming Wu, Qi Chen, Wei Yuan, Xiaodan Shi, Xiaowei Shao, Yongwei Xu, Ryosuke Shibasaki

International Geoscience and Remote Sensing Symposium (IGARSS)　2018-July　6187-6190　2018/10/31

DOI： 10.1109/IGARSS.2018.8519049 　
An Automatic Detection Method of Mismatching Points in Remote Sensing Images Based on Graph Theory,基于图论的遥感影像误匹配点自动探测方法 Peer-reviewed

Yuan, X., Yuan, W., Chen, S.

Wuhan Daxue Xuebao (Xinxi Kexue Ban)/Geomatics and Information Science of Wuhan University　43　(12)　1854-1860　2018

DOI： 10.13203/j.whugis20180154 　

ISSN： 1671-8860
Automatic building segmentation of aerial imagery usingmulti-constraint fully convolutional networks Peer-reviewed

Wu, G., Shao, X., Guo, Z., Chen, Q., Yuan, W., Shi, X., Xu, Y., Shibasaki, R.

Remote Sensing　10　(3)　407-407　2018
Publisher: {MDPI} {AG}
DOI： 10.3390/rs10030407 　

ISSN： 2072-4292
Matching multi-sensor remote sensing images via an affinity tensor Peer-reviewed

Chen, S., Yuan, X., Yuan, W., Niu, J., Xu, F., Zhang, Y.

Remote Sensing　10　(7)　2018

DOI： 10.3390/rs10071104 　

eISSN： 2072-4292
Poor textural image tie point matching via graph theory Peer-reviewed

Yuan, X., Chen, S., Yuan, W., Cai, Y.

ISPRS Journal of Photogrammetry and Remote Sensing　129　21-31　2017

DOI： 10.1016/j.isprsjprs.2017.04.015 　

ISSN： 0924-2716
Optimal seamline detection for orthoimage mosaicking by combining deep convolutional neural network and graph cuts Peer-reviewed

Li, L., Yao, J., Liu, Y., Yuan, W., Shi, S., Yuan, S.

Remote Sensing　9　(7)　701-701　2017
Publisher: {MDPI} {AG}
DOI： 10.3390/rs9070701 　

ISSN： 2072-4292

eISSN： 2072-4292
An aerial-image dense matching approach based on optical flow field Peer-reviewed

Wei Yuan, Shiyu Chen, Yong Zhang, Jianya Gong, Ryosuke Shibasaki

International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences - ISPRS Archives　41　543-548　2016

DOI： 10.5194/isprsarchives-XLI-B3-543-2016 　

ISSN： 1682-1750
Poor textural image matching based on graph theory Peer-reviewed

Shiyu Chen, Xiuxiao Yuan, Wei Yuan, Yang Cai

International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences - ISPRS Archives　41　741-747　2016

DOI： 10.5194/isprsarchives-XLI-B3-741-2016 　

ISSN： 1682-1750

Show all ︎Show first 5

Misc. 1

Exploring intercity regional similarity using worldwide location-based social network data (demo paper).

Zipei Fan, Guixu Lin, Wei Yuan, Ryosuke Shibasaki, Pengpeng E, Xuan Song 0001

SIGSPATIAL/GIS　104-4　2022

DOI： 10.1145/3557915.3561041

Presentations 22

Flood Depth Mapping from SAR Imagery Using CS-Mamba with DEM Sensitivity Analysis

Zhongyuan Yang, Wei Yuan, Weihang Ran, Changqing Liu, Bruno Adriano, Ryosuke Shibasaki, Shunichi Koshimura

ISPRS 2026　2026/07/12
Kinematic Characteristics and Risk Analysis of Potential Rockfall based on 3D Point Clouds

Wei Yuan, Changqing Liu, Han Bao, Weihang Ran, Zhongyuan Yang, Xiuxiao Yuan, Ryosuke Shibasaki, Shunichi Koshimura

ISPRS 2026　2026/07/12
From 2D Imagery to 3D Spatialtemporal Understanding Advances of Photogrammetry and Remote Sensing in Disaster Assesment and Urban Change Monitoring Invited

Wei Yuan

2025/11/12
PortVIS: An Interactive Platform for Port-to-Port Trajectory Imputation and Visual Analytics

Zhiwen Zhang, Zipei Fan, Wei Yuan, Shun Iwazaki, Ryosuke Shibasaki

SIGSPATIAL 2025　2025/11/05
You Always Recognize Me (YARM): Robust Texture Synthesis Against Multi-View Corruption

Weihang Ran, Wei Yuan, Yinqiang Zheng

Forty-second International Conference on Machine Learning　2025/07/14
Multi-source 3D point clouds fusion for potential rock mass hazard evaluation in high-steep rock slopes

Wei Yuan, Changqing Liu, Tianyi Wang, Bruno Adriano, Han Bao, Ryosuke Shibasaki, Shunichi Koshimura

ISPRS Geospatial Week 2025 Dubai　2025/04/08
The Performance of the Optical Flow Field based Dense Image Matching for UAV Imagery

Wei Yuan, Weihang Ran, Bruno Adriano, Ryosuke Shibasaki, Shunichi Koshimura

ISPRS Technical Commission IV Symposium 2024　2024/10/22
Active Polygon-based Building Outline Extraction from High-resolution Aerial Images

Weihang Ran, Wei Yuan, Zipei Fan, Xiaodan Shi, Ryosuke Shibasaki

ISPRS Geospatial Week 2023　2023/09/05
Hybrid Feature Embedding for Automatic Building Outline Extraction

Weihang Ran, Wei Yuan, Xiaodan Shi, Zipei Fan, Ryosuke Shibasaki

IGARSS 2023-2023 IEEE International Geoscience and Remote Sensing Symposium　2023/07/16
Graph Encoding based Hybrid Vision Transformer for Automatic Road Network Extraction

Wei Yuan, Weihang Ran, Xiaodan Shi, Zipei Fan, Yang Cai, Ryosuke Shibasaki

IGARSS 2023-2023 IEEE International Geoscience and Remote Sensing Symposium　2023/07/16
Few-Shot Depth Completion Using Denoising Diffusion Probabilistic Model

Weihang Ran, Wei Yuan, Ryosuke Shibasaki

IEEE/CVF Conference on Computer Vision and Pattern Recognition　2023/06
Online Trajectory Prediction for Metropolitan Scale Mobility Digital Twin

Zipei Fan, Xiaojie Yang, Wei Yuan, Renhe Jiang, Quanjun Chen, Xuan Song, Ryosuke Shibasaki

The ACM SIGSPATIAL International Conference on Advances in Geographic Information Systems 2022　2022/11/03
Exploring Intercity Regional Similarity using Worldwide Location-based Social Network Data

Zipei Fan, Guixu Lin, Wei Yuan, Ryosuke Shibasaki, Pengpeng E, Xuan Song

The ACM SIGSPATIAL International Conference on Advances in Geographic Information Systems 2022　2022/11/02
Cross-Scale Attention-based Tree Crown Detection via UAV imagery

Wei Yuan, Xiaodan Shi, Zhiling Guo, Zipei Fan, Jianya Gong, Ryosuke Shibasaki

IGARSS 2022-2022 IEEE International Geoscience and Remote Sensing Symposium　2022/07/19
Learning Social Complaint Multi-Modal Distributions of Human Path in Crowds

Xiaodan Shi, Haoran Zhang, Wei Yuan, Dou Huang, Zhiling Guo, Ryosuke Shibasaki

XXIVth ISPRS congress　2022/07/08
Graph Neural Network based Multi-Feature Fusion for Building Change Detection

Wei Yuan, Xiuxiao Yuan, Zipei Fan, Zhiling Guo, Xiaodan Shi, Jianya Gong, Ryosuke Shibasaki

XXIVth ISPRS congress　2021/07/07
End-to-end Building Change Detection Model in Aerial Imagery and Digital Surface Model based on Neural Networks

Xinlei Lian, Wei Yuan, Zhiling Guo, Zekun Cai, Xuan Song, Ryosuke Shibasaki

XXIVth ISPRS Congress　2020/09/01
Unsupervised Multi-Constraint Deep Neural Network for Dense Image Matching

Wei Yuan, Zipei Fan, Xiuxiao Yuan, Jianya Gong, Ryosuke Shibasaki

XXIVth ISPRS Congress　2020/09/02
Multimodal Interaction-aware Trajectory Prediction in Crowded Space

Xiaodan Shi, Xiaowei Shao, Zipei Fan, Renhe Jiang, Haoran Zhang, Zhiling Guo, Guangming Wu, Wei Yuan, Ryosuke Shibasaki

The Thirty-Fourth AAAI Conference on Artificial Intelligence　2020/02/10
Semantic Segmentation for Urban Planning Maps based on U-Net

Zhiling Guo, Hiroaki Shengoku, Guangming Wu, Qi Chen, Wei Yuan, Xiaodan Shi, Xiaowei Shao, Yongwei Xu, Ryosuke Shibasaki

IGARSS 2018-2018 IEEE International Geoscience and Remote Sensing Symposium　2018/07/25
Poor Textural Image Matching based on Graph Theory

Shiyu Chen, Xiuxiao Yuan, Wei Yuan, Yang Cai

XXIIIth ISPRS Congress　2016/07/15
An Aerial-Image Dense Matching Approach based on Optical Flow Field

Wei Yuan, Shiyu Chen, Yong Zhang, Jianya Gong, Ryosuke Shibasaki

2016/07/14

Show all Show first 5

Research Projects 11

Perception and Rapid 3D Gaussian Splatting for Quantitative Urban Disaster Assessment

Yuan Wei

Offer Organization: Ministry of Education, Culture, Sports, Science and Technology

System: Strategic Professional Development Program for Young Researchers

Category: TI-FRIS

Institution: Tohoku University

2026/07 - 2031/03
AI-Agent-Assisted Construction of a Standardized Multimodal Dataset for Flood Depth Prediction

Yuan Wei

Offer Organization: Ministry of Education, Culture, Sports, Science and Technology

System: AI for Science

Category: SPReAD 1000

Institution: Tohoku University

2026/09 - 2027/03
Multi-task learning based post-disaster mapping via multi-modal remote sensing observations

袁巍

Offer Organization: 日本学術振興会

System: 科学研究費助成事業若手研究

Category: 若手研究

Institution: 東京大学

2023/04/01 - 2026/03/31
Crowd Prediction and Simulation in Disaster Scenarios Based on Geographical Foundation Models

Yao Yao, Wei Yuan

Offer Organization: International Research Institute of Disaster Science， Tohoku University

System: Disaster Resilience Co-Creation Research Project

2025/06 - 2026/03

More details Close

This project focuses on crowd flow prediction in disaster scenarios, investigating: (1) Build a dynamic crowd mobility model, integrating geographic, trajectory, and digital twin data sources, to analyze the impact of extreme conditions; (2) Mechanism analysis of disaster impacts and geographical constraints on crowd response within simulated disaster scenarios; (3) State transition forecasting with geographical constraints, leveraging virtual simulation for enhanced prediction and early warning capabilities; (4) Digital twin-enabled evacuation planning and resilience optimization, validating strategies and informing policy recommendations through scenario-based insights.
Typhoon-Focused Maritime Disaster Data Platform

Zhiwen Zhang, Wei Yuan

Offer Organization: International Research Institute of Disaster Science， Tohoku University

System: Disaster Resilience Co-Creation Research Project

2025/06 - 2026/03

More details Close

This research proposes a framework for developing a multi-sourced maritime disaster data platform focused on typhoon events. The framework includes: 1) integrating satellite imagery, AIS signals, and weather data; 2) analyzing ship flow disruptions across ports before, during, and after typhoons; and 3) constructing a maritime disaster knowledge graph for decision support. The proposed data platform will support emergency decisionmaking and recovery efforts in maritime systems.
Citywide Digital Twin System for Human Moblity Simulation and Prediction under Extreme Weather

Wei Yuan, Zipei Fan

Offer Organization: International Research Institute of Disaster Science， Tohoku University

System: Disaster Resilience Co-Creation Research Project

2025/06 - 2026/03

More details Close

This research develops a comprehensive framework integrating large-scale multi-source data and advancedmodeling for digital twin-based predictions under extreme weather conditions. The framework includes: 1)Assessing the impacts of extreme weather on human mobility through spatial-temporal analysis andknowledge discovery; 2) Simulating the spread and recovery dynamics of road disruptions in urbannetworks; 3) Developing an ensemble model to predict and visualize urban traffic flows under disasterscenarios. The methodology is applied to real-world scenarios in Japan, providing practical insights intoenhancing urban resilience during extreme weather events.
Resilience patterns of multiscale human mobility under extreme rainfall events using massive individual trajectory data

Yao Yao, Wei Yuan

Offer Organization: International Research Institute of Disaster Science， Tohoku University

System: Disaster Resilience Co-Creation Research Project

2024/06 - 2025/03

More details Close

This research presents a framework utilizing massive individual trajectory data to dissect resilience patterns of human mobility across scales. The framework includes the following components: 1) Quantifying human mobility and resilience levels. 2) Extracting resilience patterns and their spatial heterogeneity. 3) Explaining differences in resilience patterns through social factors. This study will be analyzed in the context of Japan experiencing an extreme rainstorm event, with the Japanese metropolitan area as the study area.
Foundation modal based multi-modal data fusion for efficient disaster response

Wei Yuan, Zipei Fan

Offer Organization: International Research Institute of Disaster Science， Tohoku University

System: Disaster Resilience Co-Creation Research Project

2024/06 - 2025/03

More details Close

Disaster response demands accurate, prompt, and context-sensitive decisions to mitigate loss of life and property. Traditional systems may lack the nuanced understanding needed to address built environment changes, especially local critical infrastructures, such as bridges, roads, power grids, and power supplies, which are vital lifelines ensuring essential services and stability. This research seeks to propose a novel disaster-aware multimodal AI framework that assimilates different sources of data (e.g. image, weather observation social media, and aggregated mobility flows) to enable the extraction of real-time and localized knowledge about critical infrastructure conditions and effects on local communities. This research will use Japanese metropolitan area as the target area and analyzed in the context of experiencing a Typhoon condition.
Heterogeneous Graph Neural Network based Federated Mobile Crowdsensing

FAN ZIPEI, YUAN WEI

Offer Organization: Japan Society for the Promotion of Science

System: Grants-in-Aid for Scientific Research Grant-in-Aid for Scientific Research (B)

Category: Grant-in-Aid for Scientific Research (B)

Institution: The University of Tokyo

2022/04 - 2025/03
A study of satellite-based mapping of real estate price and its applications

Shibasaki Ryosuke, Seetharam KE, Wei Yuan

Offer Organization: Japan Society for the Promotion of Science

System: Grants-in-Aid for Scientific Research Grant-in-Aid for Scientific Research (B)

Category: Grant-in-Aid for Scientific Research (B)

Institution: The University of Tokyo

2019/04 - 2022/03

More details Close

Since the Covid-19 has caused major obstacles to the collection of urban land price data in developing countries, we have focused on the advancement of satellite image analysis to obtain detailed urban environmental information from satellite images and clarified the relationship with domestic land price data. We have developed advanced techniques for extracting buildings and detecting the changes using high-resolution imagery, as well as those for automatically extracting slum areas (i.e., areas with significantly low environmental quality and land prices) using mainly medium-resolution satellite imagery. These methods were found to achieve a significant improvement in accuracy. We also obtained a large amount of available land price data in Japan and analyzed the relationship with environmental factors obtained from satellite images. From these results, we have obtained a prospect to estimate land prices in large cities in developing countries from satellite images.
Developing a Mobile Phone Network Data-Driven Methodology to Quantify Community Resilience in Disaster Affected Areas

Shohei Nagata, Cynthia Chen, Erick Mas, Wei Yuan, Shunichi Koshimura, Lyra Chen

Offer Organization: Tohoku University

System: Tohoku University-University of Washington Strategic Partner Fund 2024-25

Show all Show first 5

Academic Activities 5

IGARSS 2026

2026/08/09 - 2026/08/14

Activity type: Competition, symposium, etc.
IGARSS 2025

2025/08/03 - 2025/08/08

Activity type: Competition, symposium, etc.
IGARSS 2024

2024/07/07 - 2024/07/12

Activity type: Competition, symposium, etc.
ISPRS geospatial week

Activity type: Competition, symposium, etc.
IGARSS 2023

Activity type: Peer review