Dongzi Xie
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Dongzi Xie has a total of 13 co-authors. Here are the top co-authors by number of shared publications:
- Baoshan Wang
School of Earth and Space Sciences, University of Science and Technology of China, Hefei, China
1 shared publication - Heting Hong
School of Electrical Engineering and Automation, Hefei University of Technology, Hefei, China
1 shared publication Hui Gao
School of Earth and Space Sciences, University of Science and Technology of China, Hefei, 230026 China
1 shared publication- Jianwei Ma
School of Earth and Space Sciences, Peking University, Beijing, China
1 shared publication - Jiarun Yang
Anhui University of Science & Technology
1 shared publication - Jie Zhang
School of Management, University of Science and Technology of China, Hefei 230026, China
1 shared publication
Latest Publications
- Top Cited Papers
- 26 March 2025
- Published by Seismological Society of America (SSA) in Seismological Research Letters
- Vol. 96 (4) , 2477-2488
Intelligent Traffic Monitoring with Distributed Acoustic Sensing
Distributed acoustic sensing (DAS) is promising for traffic monitoring, but its extensive data and sensitivity to vibrations, causing noise, pose computational challenges. To address this, we propose a two-step deep learning workflow with high efficiency and noise immunity for DAS-based traffic monitoring, focusing on instance vehicle trajectory segmentation and velocity estimation. Our approach begins by generating a diverse synthetic DAS dataset with labeled vehicle signals, tackling the issue of missing training labels in this field. This dataset is used to train a convolutional neural network (CNN) to detect linear vehicle trajectories from the noisy DAS data in the time–space domain. However, due to significant noise, these trajectories are often fragmented and incomplete. To enhance accuracy, we introduce a second step involving the Hough transform. This converts detected linear features into point-like energy clusters in the Hough domain. Another CNN is then employed to focus on these energies, identifying the most significant points. The inverse Hough transform is applied to these points to reconstruct complete, distinct, and noise-free linear vehicle trajectories in the time–space domain. The Hough transform plays a crucial role by enforcing a local linearity constraint on the trajectories, enhancing continuity and noise immunity, and facilitating the separation of individual trajectories and estimation of vehicle velocities (indicated by trajectory slopes in the Hough domain). Our method has shown effectiveness in real-world datasets, proving its value in real-time processing of DAS data and applicability in similar traffic monitoring scenarios. All related codes and data are available in the Data and Resources section.
- 26 March 2025
- Published by Seismological Society of America (SSA) in Seismological Research Letters
- Vol. 96 (4) , 2477-2488
Top Cited PapersIntelligent Traffic Monitoring with Distributed Acoustic Sensing
Distributed acoustic sensing (DAS) is promising for traffic monitoring, but its extensive data and sensitivity to vibrations, causing noise, pose computational challenges. To address this, we propose a two-step deep learning workflow with high efficiency and noise immunity for DAS-based traffic monitoring, focusing on instance vehicle trajectory segmentation and velocity estimation. Our approach begins by generating a diverse synthetic DAS dataset with labeled vehicle signals, tackling the issue of missing training labels in this field. This dataset is used to train a convolutional neural network (CNN) to detect linear vehicle trajectories from the noisy DAS data in the time–space domain. However, due to significant noise, these trajectories are often fragmented and incomplete. To enhance accuracy, we introduce a second step involving the Hough transform. This converts detected linear features into point-like energy clusters in the Hough domain. Another CNN is then employed to focus on these energies, identifying the most significant points. The inverse Hough transform is applied to these points to reconstruct complete, distinct, and noise-free linear vehicle trajectories in the time–space domain. The Hough transform plays a crucial role by enforcing a local linearity constraint on the trajectories, enhancing continuity and noise immunity, and facilitating the separation of individual trajectories and estimation of vehicle velocities (indicated by trajectory slopes in the Hough domain). Our method has shown effectiveness in real-world datasets, proving its value in real-time processing of DAS data and applicability in similar traffic monitoring scenarios. All related codes and data are available in the Data and Resources section.
- Top Cited PapersOpen Access
- 6 June 2023
- Published by National Academy of Sciences in Proceedings of the National Academy of Sciences
- Vol. 120 (23)
Sensing prior constraints in deep neural networks for solving exploration geophysical problems
One of the key objectives in geophysics is to characterize the subsurface through the process of analyzing and interpreting geophysical field data that are typically acquired at the surface. Data-driven deep learning methods have enormous potential for accelerating and simplifying the process but also face many challenges, including poor generalizability, weak interpretability, and physical inconsistency. We present three strategies for imposing domain knowledge constraints on deep neural networks (DNNs) to help address these challenges. The first strategy is to integrate constraints into data by generating synthetic training datasets through geological and geophysical forward modeling and properly encoding prior knowledge as part of the input fed into the DNNs. The second strategy is to design nontrainable custom layers of physical operators and preconditioners in the DNN architecture to modify or shape feature maps calculated within the network to make them consistent with the prior knowledge. The final strategy is to implement prior geological information and geophysical laws as regularization terms in loss functions for training the DNNs. We discuss the implementation of these strategies in detail and demonstrate their effectiveness by applying them to geophysical data processing, imaging, interpretation, and subsurface model building.
- 6 June 2023
- Published by National Academy of Sciences in Proceedings of the National Academy of Sciences
- Vol. 120 (23)
Top Cited PapersOpen AccessSensing prior constraints in deep neural networks for solving exploration geophysical problems
One of the key objectives in geophysics is to characterize the subsurface through the process of analyzing and interpreting geophysical field data that are typically acquired at the surface. Data-driven deep learning methods have enormous potential for accelerating and simplifying the process but also face many challenges, including poor generalizability, weak interpretability, and physical inconsistency. We present three strategies for imposing domain knowledge constraints on deep neural networks (DNNs) to help address these challenges. The first strategy is to integrate constraints into data by generating synthetic training datasets through geological and geophysical forward modeling and properly encoding prior knowledge as part of the input fed into the DNNs. The second strategy is to design nontrainable custom layers of physical operators and preconditioners in the DNN architecture to modify or shape feature maps calculated within the network to make them consistent with the prior knowledge. The final strategy is to implement prior geological information and geophysical laws as regularization terms in loss functions for training the DNNs. We discuss the implementation of these strategies in detail and demonstrate their effectiveness by applying them to geophysical data processing, imaging, interpretation, and subsurface model building.