Xiangdong Yan

2022.09--2025.12, Ph.D. in the Department of Civil and Environmental Engineering

        Swanson School of Engineering, University of Pittsburgh, USA

2020.10--2022.08, M.Eng. in the Department of Civil and Architectural Engineering

        Kyushu Institute of Technology, Japan

2019.03--2020.04, Exchange Program in the Department of Civil Engineering

        Kyushu Institute of Technology, Japan

2016.09--2020.07, B.Eng. in Department of Civil Engineering

        Henan University of Science and Technology, China

March 2026 – Present, Assistant Professor, Master's Thesis Advisor

        School of Geospatial Artificial Intelligence, East China Normal University (ECNU), Shanghai, China

        Part-time: Assistant Researcher, Embodied Intelligence and Autonomous Mapping Research Center, Hinton Spatiotemporal Intelligence Institute (Academician Jonathan Li's team)

        Responsibilities: Graduate student supervision; research in computer vision, 3D reconstruction, digital twin modeling, and embodied intelligence; scheduled to teach undergraduate Computer Vision (Fall 2026)

September 2022 – December 2025, Graduate Research Assistant

        Dept. of Civil and Environmental Engineering, Swanson School of Engineering, University of Pittsburgh, USA

August 2025 – December 2025, Teaching Assistant — Digitalization in Civil Engineering (CEE 2713)

        Swanson School of Engineering, University of Pittsburgh, USA

August 2022 – December 2022, Teaching Assistant — Introduction to Structural Analysis (CEE 1330)

        Swanson School of Engineering, University of Pittsburgh, USA

October 2020 – August 2022, Graduate Research Assistant

        Dept. of Civil and Architectural Engineering, Kyushu Institute of Technology, Japan

Dr. Xiangdong Yan is an Assistant Professor and Master's thesis advisor at the School of Geospatial Artificial Intelligence (SGAI), East China Normal University (ECNU), and a part-time Assistant Researcher at the Embodied Intelligence and Autonomous Mapping Research Center, Hinton Spatiotemporal Intelligence Institute (Academician Jonathan Li's team). He received his Ph.D. in Civil and Environmental Engineering from the University of Pittsburgh (USA) in 2025.12. (Advisor: Prof. Alessandro Fascetti). He previously obtained his M.Eng. from Kyushu Institute of Technology, Japan (advisor: Prof. Pei-Shan Chen), and his B.Eng. in Civil Engineering from Henan University of Science and Technology, China. He is proficient in Chinese, English, and Japanese. He joined ECNU SGAI in March 2026.

His research follows a Perception, Computation, Digital Twin, Execution & Verification four-layer framework. During his PhD, anchored in three large-scale engineering projects, Pennsylvania transportation infrastructure inspection (Pennsylvania State Government), climate-responsive building digital twins (NSF, $1.64M total), and intelligent slipform paving quality assessment (PennDOT), he established systematic expertise across LiDAR point cloud semantic segmentation and engineering accuracy validation, binocular stereo vision-based dynamic 3D reconstruction, photometric stereo fused with Vision Transformers for precision surface sensing, 3DGS-based high-fidelity building scene reconstruction, and multi-modal sensor fusion calibration. This work produced a complete technical pipeline spanning data acquisition, intelligent interpretation, 3D scene reconstruction, and accuracy verification, forming a mutually reinforcing framework of theory, algorithms, and engineering practice.

Looking forward, the research group will focus on the following directions:

(1) Active-Passive Fusion for Precision Inspection and Intelligent Diagnostics, integrating active sensing (LiDAR, structured light) with passive sensing (RGB cameras, thermal infrared) to build high-accuracy defect detection, damage recognition, and intelligent diagnostic systems for engineering structures and infrastructure, investigating complementary fusion mechanisms across geometric, radiometric, and thermodynamic modalities;

(2) Neural 3D Reconstruction and Engineering Digital Twin, advancing the geometric accuracy of 3DGS and NeRF frameworks for multi-resolution engineering scene reconstruction, developing sparse LiDAR point cloud-driven Gaussian initialization strategies, and enabling automated point cloud-to-digital-twin pipelines for large-scale building and infrastructure scenarios;

(3) VLM-Driven Open-Vocabulary Engineering Perception, leveraging large-scale vision-language models (CLIP, Grounded-SAM, large multimodal models) to enable open-vocabulary instance segmentation and semantic understanding of engineering scenes, overcoming the closed-category limitations of traditional supervised approaches;

(4) Embodied Intelligence and Intelligent Construction, advancing perception-cognition-decision integrated closed-loop systems for robotic construction (UR10e) and autonomous site inspection (AgileX Scout 2.0), with real-time 3D perception deployed on NVIDIA Jetson AGX Orin edge AI platforms.

Research outputs are targeted at CVPR, ICCV, ECCV, 3DV (China), ICRA, Automation in Construction, ISARC, ASCE Computing in Civil Engineering (CCE), and ACM MM.

Interested applicants: please send CV (with transcripts, research experience, and representative work) to: xdyan@geoai.ecnu.edu.cn

(1) Active-Passive Fusion for Precision Inspection and Intelligent Diagnostics

Integration of active sensing (LiDAR, structured light) and passive sensing (RGB cameras, thermal infrared) for high-accuracy defect detection and intelligent diagnostic systems targeting engineering structures and infrastructure. Research focuses on complementary fusion across geometric, radiometric, and thermodynamic modalities; deep learning-based damage segmentation and quantification; and physics-informed diagnostic reasoning.

(2) Neural 3D Reconstruction and Engineering Digital Twin

Application of 3D Gaussian Splatting (3DGS), NeRF, and extensions to multi-resolution engineering scene reconstruction. Key research: sparse LiDAR point cloud-driven Gaussian initialization for absolute-scale anchoring; multi-scale geometric constraints for accuracy; automated point cloud-to-BIM-to-digital-twin pipelines for building and infrastructure scenarios.

(3) Vision-Language Model-Driven Open-Vocabulary Engineering Perception

Leveraging large-scale VLMs (large multimodal models) for open-vocabulary instance segmentation, zero-shot material classification, and automated semantic annotation, overcoming closed-category limitations in engineering contexts. Applications include 3D point cloud semantic understanding and drawing-to-reality comparison.

(4) Embodied Intelligence and Intelligent Construction

Perception-cognition-decision integrated closed-loop systems for robotic construction and autonomous site inspection. Research covers real-time 3D sensing on edge AI platforms; high-precision localization in GNSS-denied environments; and multi-robot collaborative quality inspection.

(5) Computer Vision and Multi-Source 3D Perception

Theoretical and methodological research on CNNs, Vision Transformers, and diffusion models for 3D point cloud processing, monocular/stereo depth estimation, and photometric stereo 3D reconstruction. Focus on geometric-photometric consistency and generalizable 3D perception algorithms.

Computer Vision (Undergraduate, 3rd year; scheduled Fall 2026)

2025   NSF I-Corps Regional Program (U.S. National Science Foundation)

       Recognized for industry translation potential; admitted to the I-Corps commercialization training program

2024   CEE Travel Grant (Swanson School of Engineering, University of Pittsburgh)

2024   Research Assistant of the Year Award

       Swanson School of Engineering, University of Pittsburgh

2024   TRB Best Paper Award (AKC50)

       Transportation Research Board (TRB), USA

2022   JASSO Scholarship (Japan Student Services Organization)

2020   JASSO Scholarship (Japan Student Services Organization)

2018   National Scholarship (Ministry of Education, China)

2018   Excellence Award, National Revit Competition (Ministry of Industry and Information Technology, China)