GYGuidong Yang

Guidong Yang

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Guidong Yang has a total of 30 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • State-Based Reinforcement Learning Safe Trajectory Tracking for Autonomous Vehicles via Geometric Safety Representation and Lagrangian-PPO

    by Qi Zhang, Xiaoke Yang, Jin Chen, Yafei Wang, Guidong Yang, Ming Li

    This paper investigates a state-based Safe Trajectory Tracking (STT) control framework for autonomous vehicles in dynamic environments, where the objective is to design a controller that ensures accurate trajectory tracking while guaranteeing nonconservative collision avoidance. Existing learning-based approaches often suffer from inaccurate geometric safety representations and unstable training when incorporating safety constraints. To address these challenges, we propose a geometry-aware framework by integrating an analytical Rotating Supporting Hyperplane (RSH)-based safety representation with a Lagrangian Proximal Policy Optimization (Lagrangian-PPO) reinforcement learning algorithm. By modeling vehicles and obstacles as rigid ellipsoids, an explicitly differentiable signed distance is derived to provide a continuous and nonconservative safety cost. Based on this representation, the trajectory tracking problem is formulated as a constrained Markov decision process, enabling unified optimization of tracking performance and safety via a reward–cost decoupled learning structure. Finally, asymptotic convergence and effectiveness of the proposed approach are validated through simulation and Vehicle-in-the-Loop (ViL) experiments, demonstrating superior tracking accuracy and robust obstacle avoidance compared with existing methods.

  • Toward End-to-End Underwater Multi-View Stereo for Real-World Dense Scene Reconstruction

    by Guidong Yang, Junjie Wen, Lei Lei, Benyun Zhao, Qingxiang Li, Xi Chen, Zhi Gao, Ben M. Chen

    Multi-view stereo (MVS) enables accurate and complete 3D reconstruction from multi-view imagery, serving as a core methodology in remote sensing applications across terrestrial and underwater domains. Recent advancements in learning-based MVS have demonstrated significant improvements over traditional counterparts, primarily due to the extensive availability of multi-view training images with ground-truth metric depths in the terrestrial in-air domain. However, underwater multi-view stereo (UwMVS) faces substantial challenges arising from the domain gap between in-air and underwater environments, leading to degraded performance when applying in-air MVS models to underwater scenarios. Furthermore, the progress of learning-based UwMVS methods has been hindered by the scarcity of underwater multi-view images with ground-truth depth maps and point clouds. In this paper, we address these challenges by introducing a physically-guided approach for synthesizing underwater multi-view images and presenting the first large-scale synthetic UwMVS dataset preserving real-world underwater degradation properties for end-to-end training and evaluation of learning-based UwMVS methods. Furthermore, we propose a novel UwMVS network that enhances geometric cue encoding to achieve more accurate and complete point cloud reconstruction. Extensive experiments on the dataset and real-world underwater scenes demonstrate that our dataset enables the trained models for underwater dense reconstruction and that our method achieves state-of-the-art performance in underwater reconstruction. Dataset, appendix, and supplementary video are available at https://yang-sober.github.io/UnderMVS/.


  • Multi-View Stereo With Geometric Encoding for Large-Scale Dense Scene Reconstruction

    by Guidong Yang, Rui Cao, Junjie Wen, Benyun Zhao, Qingxiang Li, Xi Chen, Yun-Hui Liu, Ben M. Chen

    Multi-view stereo (MVS) implicitly encodes photometric and geometric cues into the cost volume for multi-view correspondence matching, transferring insufficient geometric cues essential to depth estimation and reconstruction. This paper proposes GE-MVS, a novel multi-view stereo network with geometric encoding for more accurate and complete depth estimation and point cloud reconstruction. First, the cross-view adaptive cost volume aggregation module is proposed to strengthen the encoding of multi-view geometric cues during cost volume construction. Then, the depth consistency optimization is performed in 3D point space during learning by invoking ground-truth depth cues from adjacent views. Finally, the surface normal geometries are explicitly encoded to refine the sampled depth hypotheses to be consistent in the local neighbor regions. Extensive experiments on the standard MVS benchmarks including DTU, Tanks and Temples, and BlendedMVS demonstrate the state-of-the-art depth estimation and point cloud reconstruction performance of GE-MVS. The GE-MVS is further deployed in real-world experiments for UAV-based large-scale reconstruction, where our method outperforms the prevalent industrial reconstruction solutions in terms of reconstruction efficiency and effectiveness. Supplementary video can be found at https://youtu.be/Z4tGROatVjU.


  • Det-Recon-Reg: An Intelligent Framework Toward Automated UAV-Based Large-Scale Infrastructure Inspection

    by Guidong Yang, Benyun Zhao, Jihan Zhang, Junjie Wen, Qingxiang Li, Lei Lei, Xi Chen, Ben M. Chen

    Visual inspection remains essential for inspecting infrastructure surfaces. While there are cornerstones in developing intelligent inspection systems, most existing solutions are limited to small-scale infrastructures and components, making them challenging to scale up for real-world applications. Leveraging deep learning and unmanned aerial vehicles, this paper proposes Det-Recon-Reg, an intelligent framework born for large-scale infrastructure inspection by decomposing it into three complementary stages: Detect for defect detection, Reconstruct for infrastructure reconstruction, and Register for defect registration. In the Detect stage, we introduce the first highresolution dataset designed for defect detection on large-scale infrastructure surfaces. State-of-the-art real-time object detectors are evaluated on this dataset and the CUBIT-Net1 is proposed to strike a better balance between accuracy and efficiency. In the Reconstruct stage, we present a scalable multi-view stereo (MVS) network to reconstruct dense point cloud representation of the infrastructure from multi-view images. Extensive experiments on benchmark datasets, including DTU, Tanks and Temples, and BlendedMVS, demonstrate the superior performance of our method over existing approaches. In the Register stage, we propose a novel defect registration method that leverages the geographic information system to accurately map the detected defects onto the infrastructure model while preserving their geometric and visual properties, thereby enabling global defect localization and more informed maintenance decision-making. The proposed framework can serve as a reference for effective and efficient infrastructure maintenance as consolidated in real-world experiments. Codes, datasets, and pre-trained models for each stage will be released at: https://github.com/YANG-SOBER/Det-Recon-Reg. The supplementary video is available at: https://youtu.be/MVMp7k9qB84.


  • Sparse-to-Dense Prediction of Ocean Subsurface Temperature Using Multilevel Spatiotemporal Information Fusion

    by Lei Lei, Guidong Yang, Zuoquan Zhao, Xi Chen, Ben M. Chen

    Accurately predicting ocean subsurface temperature is vital for advancing ocean and climate research, particularly given the sparse and costly nature of subsurface observations. This study introduces sparse-to-dense prediction of ocean subsurface temperature using multi-level spatiotemporal (ST) information fusion. The framework integrates interpretable ST decoupling, adaptive feature updating, and sparse-to-dense information fusion modules to address the challenge of sparse observations and ever-evolving dynamic environments. Comprehensive experiments focused on the Pacific demonstrate the superiority of the proposed methodology over peer methods. The proposed methodology achieves high-resolution predictions with a root mean square error of 0.2230, accuracy of 0.9846, and point-wise prediction errors below 0.5°C under 10% online random sparse observations (ORSO). Analyses of spatial and temporal temperature dynamics reveal long-term warming trends in the Pacific, including a temperature rise of up to 2.8°C at -100 m in low-latitude regions over the past 40 years, and identify the latitudinal slope of thermocline dynamics. This study advances the understanding of multi-scale thermal processes and variability in the Pacific, demonstrating the potential of application in climate studies, marine resource management, and environmental monitoring.


  • WaterFormer: A Global–Local Transformer for Underwater Image Enhancement With Environment Adaptor

    by Junjie Wen, Jinqiang Cui, Guidong Yang, Benyun Zhao, Yu Zhai, Zhi Gao, Lihua Dou, Ben M. Chen

    Underwater image enhancement (UIE) is crucial for high-level vision in underwater robotics. While convolutional neural networks (CNNs) have made significant achievements in UIE, the locality of convolution poses a challenge in capturing the global context. In contrast, transformer-based networks, adept at handling long-range dependencies, have shown promise in various vision tasks. Nonetheless, directly applying a transformer to UIE faces critical challenges: 1) it tends to produce results with coarse details due to the negligence of local texture and 2) the varicolored degraded images require the network to be adaptable to different underwater environments. In this article, we propose a novel transformer-based network that can effectively leverage both the global contextual and local detailed information with some key designs (a global–local transformer [GL-Trans] block and a detail-enhanced skip connector [DESC]) while being computationally efficient. Moreover, by introducing a simple but effective learnable environment adaptor, the proposed network is flexible to deal with different underwater environments. Extensive experiments have been conducted and have demonstrated the superiority of our proposed network compared with other state-of-the-art (SOTA) methods both qualitatively and quantitatively.


  • VDTNet: A High-Performance Visual Network for Detecting and Tracking of Intruding Drones

    by Xunkuai Zhou, Guidong Yang, Yizhou Chen, Li Li, Ben M. Chen

    The misuse of drones can jeopardize public safety and privacy. The detection and catching of intruding drones are crucial and urgent issues to be investigated. This work proposes VDTNet, an accurate, lightweight, and fast network for visually detecting and tracking intruding drones. We first incorporate an SPP module into the first head of YOLOv4 to enhance detection accuracy. Model compression is utilized to shrink the model size and concurrently speed up inference. We then propose and insert an SPPS module and a ResNeck module into the neck, and introduce an effective attention module for the backbone to compensate for the accuracy drop brought on by compression. With the above strategies, we present the accurate and compact VDTNet with a model size of merely 3.9 MB, ensuring low computational cost and fast detection and tracking performance in real time. Extensive experiments on four challenging public datasets show that our proposed network outperforms state-of-the-art approaches. In real-world scenarios, the comparative ground-to-air detection testing proves the generalization ability of the VDTNet, and we further demonstrate the portability and practicability of the network by deploying it on drone onboard edge-computing devices for air-to-air real-time detection of the intruding drones.


  • Learnable Cost Metric-Based Multi-View Stereo for Point Cloud Reconstruction

    by Guidong Yang, Xunkuai Zhou, Chuanxiang Gao, Xi Chen, Ben M. Chen

    3-D reconstruction is essential to defect localization. This article proposes LCM-MVSNet, a novel multi-view stereo (MVS) network with learnable cost metric (LCM) for more accurate and complete dense point cloud reconstruction. To adapt to the scene variation and improve the reconstruction quality in non-Lambertian low-textured scenes, we propose LCM to adaptively aggregate multi-view matching similarity into the 3-D cost volume by leveraging sparse point hints. The proposed LCM benefits the MVS approaches in four folds, including depth estimation enhancement, reconstruction quality improvement, memory footprint reduction, and computational burden alleviation, allowing the depth inference for high-resolution images to achieve more accurate and complete reconstruction. In addition, we improve the depth estimation by enhancing the shallow feature propagation via a bottom–up pathway and strengthen the end-to-end supervision by adapting the focal loss to reduce ambiguity caused by sample imbalance. Extensive experiments on three benchmark datasets show that our method achieves state-of-the-art performance on the DTU and BlendedMVS dataset, and exhibits strong generalization ability with a competitive performance on the Tanks and Temples benchmark. Furthermore, we deploy our LCM-MVSNet into our UAV-based infrastructure defect inspection framework for infrastructure reconstruction and defect localization, demonstrating the effectiveness and efficiency of our method. More experiment results can be found in the Appendix at https://github.com/CUHK-USR-Group/TIE_Appendices/blob/main/TIE_Appendix.pdf .