Xiaojun Lu
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Xiaojun Lu has a total of 8 co-authors. Here are the top co-authors by number of shared publications:
- Atsushi Yamashita
The University of Tokyo
2 shared publications - Yongdong Wang
Okayama University
2 shared publications - Angela Faragasso
Department of Precision Engineering, The University of Tokyo, Tokyo, Japan
1 shared publication Hajime Asama
The University of Tokyo
1 shared publication- Hajime Asama
The University of Tokyo
1 shared publication Linchu Yang
Jiangsu University of Science and Technology
1 shared publication
Latest Publications
- 10 November 2026
- Published by IOP Publishing in Measurement Science and Technology
A high-precision measurement method for dynamic cable winding gaps in multi-layer winding winches based on machine vision
High-precision online detection of cable winding gaps in tethered balloon winches is critical to ensuring the safe and stable operation of tethered systems. To address perspective distortion caused by the cylindrical drum surface, spurious edge interference from cable twisted texture and insufficient measurement accuracy under dynamic multi-layer winding conditions, this paper proposes a vision-based cable gap measurement method combining multi-layer geometric modelling with local directional dispersion enhancement. A recursive projection model of wave-like curves for multi-layer cables is constructed, which compensates for perspective errors from curved-surface depth variations through layer-by-layer object-distance correction. A Devernay-LoDD sub-pixel edge detection algorithm is proposed, which jointly suppresses texture spurious edges using gradient peak confidence and geometric direction confidence. Combined with dynamic ROI localisation, exponentially weighted guided filtering (EWGF) and RANSAC line fitting, a complete measurement pipeline is established. A visual detection system is built for experimental verification. The results show that the standard deviation for static repeated measurements is less than 0.004 mm, with a maximum absolute error of 0.0159 mm. Under the strong texture condition of the second layer, the mean absolute error is reduced by 56.4% compared with the traditional Devernay algorithm. Three-layer dynamic winding tests yield an average relative error of approximately 5.09%, with absolute errors below 0.10 mm for most measurement points. The geometric model has a layer-by-layer recursive property and can be extended to more winding layers, meeting the demand for high-precision online detection of cable winding gaps in tethered balloon winches.
- 10 November 2026
- Published by IOP Publishing in Measurement Science and Technology
A high-precision measurement method for dynamic cable winding gaps in multi-layer winding winches based on machine vision
High-precision online detection of cable winding gaps in tethered balloon winches is critical to ensuring the safe and stable operation of tethered systems. To address perspective distortion caused by the cylindrical drum surface, spurious edge interference from cable twisted texture and insufficient measurement accuracy under dynamic multi-layer winding conditions, this paper proposes a vision-based cable gap measurement method combining multi-layer geometric modelling with local directional dispersion enhancement. A recursive projection model of wave-like curves for multi-layer cables is constructed, which compensates for perspective errors from curved-surface depth variations through layer-by-layer object-distance correction. A Devernay-LoDD sub-pixel edge detection algorithm is proposed, which jointly suppresses texture spurious edges using gradient peak confidence and geometric direction confidence. Combined with dynamic ROI localisation, exponentially weighted guided filtering (EWGF) and RANSAC line fitting, a complete measurement pipeline is established. A visual detection system is built for experimental verification. The results show that the standard deviation for static repeated measurements is less than 0.004 mm, with a maximum absolute error of 0.0159 mm. Under the strong texture condition of the second layer, the mean absolute error is reduced by 56.4% compared with the traditional Devernay algorithm. Three-layer dynamic winding tests yield an average relative error of approximately 5.09%, with absolute errors below 0.10 mm for most measurement points. The geometric model has a layer-by-layer recursive property and can be extended to more winding layers, meeting the demand for high-precision online detection of cable winding gaps in tethered balloon winches.
- 10 March 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Robotics and Automation Letters
- Vol. 10 (4) , 4140-4147
Group-Aware Robot Navigation in Crowds Using Spatio-Temporal Graph Attention Network With Deep Reinforcement Learning
Robots are becoming essential in human environments, requiring them to behave in a socially compliant manner. Although previous learning-based methods have shown potential in social navigation, most have treated pedestrians as individuals, failing to account for group level interactions. Additionally, these methods have modeled pairwise interactions only in the spatial domain, overlooking the temporal evolution of relations among agents. In this letter, the above limitations are addressed by proposing a novel spatio-temporal graph attention network that explicitly models group level interactions in both spatial and temporal domains. Specifically, a novel group-awareness mechanism is designed to model group-aware behaviors, and a new network is proposed to capture spatio-temporal features of relations among agents while leveraging the model-free deep reinforcement learning to optimize the group-aware navigation policy. The test results show that our approach outperforms the baselines in all metrics in both simulation and real-world experiments. Furthermore, quantitative analysis of questionnaire responses further verifies the benefits of our method in group awareness and social compliance.
- 10 March 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Robotics and Automation Letters
- Vol. 10 (4) , 4140-4147
Group-Aware Robot Navigation in Crowds Using Spatio-Temporal Graph Attention Network With Deep Reinforcement Learning
Robots are becoming essential in human environments, requiring them to behave in a socially compliant manner. Although previous learning-based methods have shown potential in social navigation, most have treated pedestrians as individuals, failing to account for group level interactions. Additionally, these methods have modeled pairwise interactions only in the spatial domain, overlooking the temporal evolution of relations among agents. In this letter, the above limitations are addressed by proposing a novel spatio-temporal graph attention network that explicitly models group level interactions in both spatial and temporal domains. Specifically, a novel group-awareness mechanism is designed to model group-aware behaviors, and a new network is proposed to capture spatio-temporal features of relations among agents while leveraging the model-free deep reinforcement learning to optimize the group-aware navigation policy. The test results show that our approach outperforms the baselines in all metrics in both simulation and real-world experiments. Furthermore, quantitative analysis of questionnaire responses further verifies the benefits of our method in group awareness and social compliance.