Lu Yang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Lu Yang has a total of 46 co-authors. Here are the top co-authors by number of shared publications:
- Chongke Bi
Tianjin University
15 shared publications - Xiaobin Qiu
Tianjin Key Laboratory for Oceanic Meteorology, Tianjin, China
6 shared publications - Ce Yu
Tianjin University, Tianjin, China
4 shared publications - Yiwen Bai
School of Transportation Science and Engineering, Beihang University, Beijing 102206, People’s Republic of China
4 shared publications - Chenyue Jiao
College of Intelligence and Computing, Tianjin University, Tianjin, China
3 shared publications - Fenglei Ren
Tianjin Key Laboratory for Advanced Mechanical System Design and Intelligent Control, School of Mechanical Engineering, Tianjin University of Technology, Tianjin, China
3 shared publications
Latest Publications
- Open Access
- 24 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 2012-2019
Dynamic Crowd Counting Monitoring System for Smart Classrooms Based on YOLOv8s and ByteTrack Multi-Object Tracking
To address the pervasive challenges of headcount inaccuracies and statistical anomalies arising from dense crowd distributions and severe inter-target occlusions in smart classrooms, this study proposes a robust dynamic monitoring framework integrating YOLOv8s and the ByteTrack algorithm. The system constructs a detection-tracking dual-coupled architecture: the lightweight YOLOv8s is adopted at the detection end to extract student features, ensuring the capability of capturing small targets in complex scenes; the ByteTrack algorithm is introduced at the tracking end to realize continuous prediction of target motion states through Kalman Filter, which effectively smooths the detection fluctuations caused by instantaneous occlusion. Experimental results based on the real classroom video stream dataset (SCB-Dataset) show that the system achieves a recognition accuracy of over 90% in dense crowd scenarios, with an extremely low ID switch rate and strong anti-occlusion performance. Furthermore, accelerated by GPU heterogeneous computing, our edge-deployed system achieves a real-time inference speed of 80 FPS, thereby establishing a low-latency technical foundation for fine-grained resource allocation and intelligent energy management within campus ecosystems.
- 24 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 2012-2019
Open AccessDynamic Crowd Counting Monitoring System for Smart Classrooms Based on YOLOv8s and ByteTrack Multi-Object Tracking
To address the pervasive challenges of headcount inaccuracies and statistical anomalies arising from dense crowd distributions and severe inter-target occlusions in smart classrooms, this study proposes a robust dynamic monitoring framework integrating YOLOv8s and the ByteTrack algorithm. The system constructs a detection-tracking dual-coupled architecture: the lightweight YOLOv8s is adopted at the detection end to extract student features, ensuring the capability of capturing small targets in complex scenes; the ByteTrack algorithm is introduced at the tracking end to realize continuous prediction of target motion states through Kalman Filter, which effectively smooths the detection fluctuations caused by instantaneous occlusion. Experimental results based on the real classroom video stream dataset (SCB-Dataset) show that the system achieves a recognition accuracy of over 90% in dense crowd scenarios, with an extremely low ID switch rate and strong anti-occlusion performance. Furthermore, accelerated by GPU heterogeneous computing, our edge-deployed system achieves a real-time inference speed of 80 FPS, thereby establishing a low-latency technical foundation for fine-grained resource allocation and intelligent energy management within campus ecosystems.
- 5 February 2026
- Published by IOP Publishing in Engineering Research Express
- Vol. 8 (4)
DC-SLAM: Dynamic Prior Update and Missed Detection Compensation for Robust RGB-D SLAM
Simultaneous Localization and Mapping (SLAM) is a fundamental technology used for robotic perception and navigation. Conventional SLAM systems assume static environments, which often fails in real-world dynamic scenarios where moving objects cause significant errors or tracking failures. Current approaches to dynamic feature filtering primarily rely on either multi-view geometry or deep learning techniques. While geometry-based approaches are computationally intensive and environmentally sensitive, deep learning-based methods are constrained by fixed priors, making them less effective at handling pseudo-static objects and prone to missed detections under occlusion or motion blur. To overcome these limitations, we present DC-SLAM, a framework that synergistically combines geometric constraints with deep learning to enable both accurate dynamic prior updates and robust missed-detection compensation in real-time operation. Its key innovations include, first, accurately identifying pseudo-static objects using a dynamic reference frame comparison algorithm and decoupling their motion from the background via depth information and multi-view geometric constraints, enabling real-time prior updates. Second, to address missed detections from sudden viewpoint changes, motion blur, or occlusions, it predicts target positions through spatiotemporal motion consistency analysis and locates compensation boxes within expanded regions via a column-wise depth clustering algorithm, greatly improving robustness. To evaluate the proposed algorithm, comprehensive experiments are conducted on the TUM RGB-D benchmark against state-of-the-art SLAM systems, along with real-world tests in dynamic environments. On four TUM RGB-D Fr3 dynamic sequences, DC-SLAM attains the lowest ATE across all methods, with 93–98% lower ATE RMSE than ORB-SLAM3 and up to 80\% lower than recent dynamic-aware baselines including DS-SLAM, RDS-SLAM, and SG-SLAM. Real-world mobile robot experiments further demonstrate DC-SLAM's capability to reliably filter dynamic objects and maintain precise localization, while exhibiting exceptional robustness against challenging conditions such as viewpoint variations and occlusions, ultimately delivering a highly dependable visual SLAM solution for dynamic environments.
- 5 February 2026
- Published by IOP Publishing in Engineering Research Express
- Vol. 8 (4)
DC-SLAM: Dynamic Prior Update and Missed Detection Compensation for Robust RGB-D SLAM
Simultaneous Localization and Mapping (SLAM) is a fundamental technology used for robotic perception and navigation. Conventional SLAM systems assume static environments, which often fails in real-world dynamic scenarios where moving objects cause significant errors or tracking failures. Current approaches to dynamic feature filtering primarily rely on either multi-view geometry or deep learning techniques. While geometry-based approaches are computationally intensive and environmentally sensitive, deep learning-based methods are constrained by fixed priors, making them less effective at handling pseudo-static objects and prone to missed detections under occlusion or motion blur. To overcome these limitations, we present DC-SLAM, a framework that synergistically combines geometric constraints with deep learning to enable both accurate dynamic prior updates and robust missed-detection compensation in real-time operation. Its key innovations include, first, accurately identifying pseudo-static objects using a dynamic reference frame comparison algorithm and decoupling their motion from the background via depth information and multi-view geometric constraints, enabling real-time prior updates. Second, to address missed detections from sudden viewpoint changes, motion blur, or occlusions, it predicts target positions through spatiotemporal motion consistency analysis and locates compensation boxes within expanded regions via a column-wise depth clustering algorithm, greatly improving robustness. To evaluate the proposed algorithm, comprehensive experiments are conducted on the TUM RGB-D benchmark against state-of-the-art SLAM systems, along with real-world tests in dynamic environments. On four TUM RGB-D Fr3 dynamic sequences, DC-SLAM attains the lowest ATE across all methods, with 93–98% lower ATE RMSE than ORB-SLAM3 and up to 80\% lower than recent dynamic-aware baselines including DS-SLAM, RDS-SLAM, and SG-SLAM. Real-world mobile robot experiments further demonstrate DC-SLAM's capability to reliably filter dynamic objects and maintain precise localization, while exhibiting exceptional robustness against challenging conditions such as viewpoint variations and occlusions, ultimately delivering a highly dependable visual SLAM solution for dynamic environments.
- 2 January 2026
- Published by Springer Nature
- p. 316-326
Construction and Optimization of an Industrial Intelligent Production Platform Based on Digital Twin Technology
- 2 January 2026
- Published by Springer Nature
- p. 316-326
Construction and Optimization of an Industrial Intelligent Production Platform Based on Digital Twin Technology
- 21 September 2025
- Published by Springer Nature in Journal of Visualization
- Vol. 28 (6) , 1103-1117
A light multimodal neural network for tropical cyclone path prediction with single time step
- 21 September 2025
- Published by Springer Nature in Journal of Visualization
- Vol. 28 (6) , 1103-1117
A light multimodal neural network for tropical cyclone path prediction with single time step
- Open Access
- 11 December 2024
- Published by Association for Computing Machinery (ACM)
- p. 1-8
D-Markov: A Sparse Sample-based Model for Interannual Precipitation Prediction during the Rainy Season
- 11 December 2024
- Published by Association for Computing Machinery (ACM)
- p. 1-8
Open AccessD-Markov: A Sparse Sample-based Model for Interannual Precipitation Prediction during the Rainy Season
- Open Access
- 22 November 2024
- Published by Association for Computing Machinery (ACM)
- p. 250-255
Project-guided Teaching Reform of Image Processing and Machine Vision: A Case Study of Intelligent Diagnosis of Pulmonary Nodules
- 22 November 2024
- Published by Association for Computing Machinery (ACM)
- p. 250-255
Open AccessProject-guided Teaching Reform of Image Processing and Machine Vision: A Case Study of Intelligent Diagnosis of Pulmonary Nodules
- Open Access
- 1 November 2024
- Published by Association for Computing Machinery (ACM)
- p. 83-88
Research on Practical Teaching System of Intelligent Manufacturing Engineering: A Case Study of Major Design
Based on the construction requirements of the "Emerging Engineering" major in intelligent manufacturing engineering and the characteristics of interdisciplinary integration, the cultivation of high-quality interdisciplinary talents has been established as the primary objective. By coordinating multiple on-campus and off-campus resources and fostering cohesion within the entire discipline supervisor team, the "academic supervisor system" training mode has been established, and the continuous training mechanism for "undergraduate and master's" programs has been effectively developed. Additionally, school–enterprise collaborations have been implemented, and practical bases have been successfully established. With the support of the entire discipline supervisor team and the closed-loop integrated laboratory, the practical teaching system characterized by cross-integration and systematic progression has been developed at our university. By creating a step-by-step, planned, and systematic practical course system, the issues of standardization, randomization, and fragmentation within practical courses have been effectively resolved. Through the integration of project-based courses and school–enterprise co-constructed courses, students' innovative abilities are further enhanced. It then describes the design and implementation methods of major design, and analyzes its practical effects. It holds significant reference and demonstrative value for the construction of practical courses, laboratories, and the talent cultivation model of the "Emerging Engineering" major.
- 1 November 2024
- Published by Association for Computing Machinery (ACM)
- p. 83-88
Open AccessResearch on Practical Teaching System of Intelligent Manufacturing Engineering: A Case Study of Major Design
Based on the construction requirements of the "Emerging Engineering" major in intelligent manufacturing engineering and the characteristics of interdisciplinary integration, the cultivation of high-quality interdisciplinary talents has been established as the primary objective. By coordinating multiple on-campus and off-campus resources and fostering cohesion within the entire discipline supervisor team, the "academic supervisor system" training mode has been established, and the continuous training mechanism for "undergraduate and master's" programs has been effectively developed. Additionally, school–enterprise collaborations have been implemented, and practical bases have been successfully established. With the support of the entire discipline supervisor team and the closed-loop integrated laboratory, the practical teaching system characterized by cross-integration and systematic progression has been developed at our university. By creating a step-by-step, planned, and systematic practical course system, the issues of standardization, randomization, and fragmentation within practical courses have been effectively resolved. Through the integration of project-based courses and school–enterprise co-constructed courses, students' innovative abilities are further enhanced. It then describes the design and implementation methods of major design, and analyzes its practical effects. It holds significant reference and demonstrative value for the construction of practical courses, laboratories, and the talent cultivation model of the "Emerging Engineering" major.
- 5 June 2024
- Published by Springer Nature in Journal of Visualization
- Vol. 27 (5) , 997-1014
BLCM: a BP-LGBM-based atmospheric visibility forecasting model
- 5 June 2024
- Published by Springer Nature in Journal of Visualization
- Vol. 27 (5) , 997-1014
BLCM: a BP-LGBM-based atmospheric visibility forecasting model
- Open Access
- 22 May 2024
- Published by Association for Computing Machinery (ACM)
- p. 83-90
Tracing-U-Net: An Attention Based U-Net model for Air Pollution Source Tracing from Sparse Dataset
In recent years, the increasingly severe air pollution has caused significant harm to us, so it is imperative to effectively manage air pollution from its source. Existing methods on pollutant tracing primarily falls into two types: numerical solution simulation and deep learning. Numerical solution simulation methods are computationally intensive and fail to meet the requirements fort tracking the source of air pollution in real time. Deep learning methods often require a large amount of sample data, however, the available meteorological observation data is exceedingly sparse. To address these issues, this paper proposes a new model, called Tracing-U-Net, for efficiently tracing the source of air pollution. Firstly, to confront the challenge of sparse meteorological data, we have proposed a U-Net based model. This is because its unique skip connection design enables it to learn and retain detail information at different resolutions. Then, in order to further improve the precision of the U-Net model, we employs channel and spatial attention modules to capture crucial features and spatial distribution patterns. Also, we have designed a propagation path-generating method based on the pollution image sequence. Each image shows the pollution density in one time step. It can assist domain scientists to interactively explore the possible sources of air pollution. Finally, in order to demonstrate the effectiveness of the proposed method, several pollutant source tracing instances have been carried out. The results show that the pollutant source can be successfully traced from the sparse meteorological observation data.
- 22 May 2024
- Published by Association for Computing Machinery (ACM)
- p. 83-90
Open AccessTracing-U-Net: An Attention Based U-Net model for Air Pollution Source Tracing from Sparse Dataset
In recent years, the increasingly severe air pollution has caused significant harm to us, so it is imperative to effectively manage air pollution from its source. Existing methods on pollutant tracing primarily falls into two types: numerical solution simulation and deep learning. Numerical solution simulation methods are computationally intensive and fail to meet the requirements fort tracking the source of air pollution in real time. Deep learning methods often require a large amount of sample data, however, the available meteorological observation data is exceedingly sparse. To address these issues, this paper proposes a new model, called Tracing-U-Net, for efficiently tracing the source of air pollution. Firstly, to confront the challenge of sparse meteorological data, we have proposed a U-Net based model. This is because its unique skip connection design enables it to learn and retain detail information at different resolutions. Then, in order to further improve the precision of the U-Net model, we employs channel and spatial attention modules to capture crucial features and spatial distribution patterns. Also, we have designed a propagation path-generating method based on the pollution image sequence. Each image shows the pollution density in one time step. It can assist domain scientists to interactively explore the possible sources of air pollution. Finally, in order to demonstrate the effectiveness of the proposed method, several pollutant source tracing instances have been carried out. The results show that the pollutant source can be successfully traced from the sparse meteorological observation data.