Guiyuan Yuan
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Guiyuan Yuan has a total of 52 co-authors. Here are the top co-authors by number of shared publications:
- Jiujun Cheng
School of Computer Science and Technology, Key Laboratory of Embedded System and Service Computing (Ministry of Education), Tongji University, Shanghai, China
12 shared publications - Shangce Gao
University of Toyama
9 shared publications - MengChu Zhou
Macau University of Science and Technology
8 shared publications - Qingtian Zeng
Shandong University of Science and Technology, Qingdao, China
8 shared publications Cong Liu
School of Computer Science and Technology, Shandong University of Technology, Zibo, China
5 shared publications- Hua Duan
College of Computer Science and Engineering, Shandong University of Science and Technology, Qingdao, China
5 shared publications
Latest Publications
- 23 February 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Circuits and Systems for Video Technology
- Vol. 36  (7) , 10794-10809
Pose-Guided Multi-Cue Explicit Query Construction for Disambiguating Human--Object Interactions
Human-Object Interaction (HOI) detection remains challenging due to the semantic ambiguity of interaction categories and the limited discriminability of their feature representations. Existing approaches often improve recognition by employing sophisticated models or auxiliary textual annotations. While effective in certain gains, these solutions incur additional computational or annotation costs and struggle to capture intrinsic interaction regularities. To address these issues, we propose Pose-Guided Multi-Cue Explicit Query Construction (PM-EQC), a unified Transformer-based framework that builds upon collaborative modeling of appearance, spatial, and pose cues for discriminative interaction reasoning. At its core, the Collaborative Multi-Cue Query Constructor (CM-CQC) jointly models dependencies among visual cues to generate explicit query embeddings. CM-CQC further incorporates a hierarchical pose contextualization mechanism: global body configurations adaptively guide attention to local critical joints, yielding fine-grained pose embeddings and more precise interaction disambiguation. Owing to its modular design, PM-EQC integrates seamlessly with diverse backbones and benefits from their advances. Extensive experiments on PhysLab, HICO-DET, and V-COCO datasets demonstrate that PM-EQC achieves state-of-the-art performance, and the code is publicly available at https://github.com/ZMHSDUST/ PM-EQC.
- 23 February 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Circuits and Systems for Video Technology
- Vol. 36  (7) , 10794-10809
Pose-Guided Multi-Cue Explicit Query Construction for Disambiguating Human--Object Interactions
Human-Object Interaction (HOI) detection remains challenging due to the semantic ambiguity of interaction categories and the limited discriminability of their feature representations. Existing approaches often improve recognition by employing sophisticated models or auxiliary textual annotations. While effective in certain gains, these solutions incur additional computational or annotation costs and struggle to capture intrinsic interaction regularities. To address these issues, we propose Pose-Guided Multi-Cue Explicit Query Construction (PM-EQC), a unified Transformer-based framework that builds upon collaborative modeling of appearance, spatial, and pose cues for discriminative interaction reasoning. At its core, the Collaborative Multi-Cue Query Constructor (CM-CQC) jointly models dependencies among visual cues to generate explicit query embeddings. CM-CQC further incorporates a hierarchical pose contextualization mechanism: global body configurations adaptively guide attention to local critical joints, yielding fine-grained pose embeddings and more precise interaction disambiguation. Owing to its modular design, PM-EQC integrates seamlessly with diverse backbones and benefits from their advances. Extensive experiments on PhysLab, HICO-DET, and V-COCO datasets demonstrate that PM-EQC achieves state-of-the-art performance, and the code is publicly available at https://github.com/ZMHSDUST/ PM-EQC.
- Open Access
- 7 February 2026
- Published by MDPI AG in Applied Sciences
- Vol. 16  (4) , 1673
Small Object Detection with Efficient Multi-Scale Collaborative Attention and Depth Feature Fusion Based on Detection Transformer
Existing DEtection TRansformer-based (DETR) object detection methods have been widely applied to standard object detection tasks, but still face numerous challenges in detecting small objects. These methods frequently miss the fine details of small objects and fail to preserve global context, particularly under scale variation or occlusion. The resulting feature maps lack sufficient spatial and structural information. Moreover, some DETR-based models specifically designed for small object detection often have poor generalization capabilities and are difficult to adapt to datasets with diverse object scales and complex backgrounds. To address these issues, this paper proposes a novel object detection model—small object detection with efficient multi-scale collaborative attention and depth feature fusion based on DETR (ED-DETR)—which consists of three core modules: an efficient multi-scale collaborative attention mechanism (EMCA), DepthPro, a zero-shot metric monocular depth estimation model, and an adaptive feature fusion module for depth maps and feature maps. Specifically, EMCA extends the single-space attention mechanism in efficient multi-scale attention (EMA) to a composite structure of parallel spatial and channel attention, enhancing ED-DETR’s ability to express features collaboratively in both spatial and channel dimensions. DepthPro generates depth maps to extract depth information. The adaptive feature fusion module integrates depth information with RGB visual features, improving ED-DETR’s ability to perceive object position, scale, and occlusion. The experimental results show that ED-DETR achieves the current best 33.6% mAP on the AI-TOD-V2 dataset, which predominantly contains tiny objects, outperforming previous CNN-based and DETR-based methods, and shows excellent generalization performance on the VisDrone and COCO datasets.
- 7 February 2026
- Published by MDPI AG in Applied Sciences
- Vol. 16  (4) , 1673
Open AccessSmall Object Detection with Efficient Multi-Scale Collaborative Attention and Depth Feature Fusion Based on Detection Transformer
Existing DEtection TRansformer-based (DETR) object detection methods have been widely applied to standard object detection tasks, but still face numerous challenges in detecting small objects. These methods frequently miss the fine details of small objects and fail to preserve global context, particularly under scale variation or occlusion. The resulting feature maps lack sufficient spatial and structural information. Moreover, some DETR-based models specifically designed for small object detection often have poor generalization capabilities and are difficult to adapt to datasets with diverse object scales and complex backgrounds. To address these issues, this paper proposes a novel object detection model—small object detection with efficient multi-scale collaborative attention and depth feature fusion based on DETR (ED-DETR)—which consists of three core modules: an efficient multi-scale collaborative attention mechanism (EMCA), DepthPro, a zero-shot metric monocular depth estimation model, and an adaptive feature fusion module for depth maps and feature maps. Specifically, EMCA extends the single-space attention mechanism in efficient multi-scale attention (EMA) to a composite structure of parallel spatial and channel attention, enhancing ED-DETR’s ability to express features collaboratively in both spatial and channel dimensions. DepthPro generates depth maps to extract depth information. The adaptive feature fusion module integrates depth information with RGB visual features, improving ED-DETR’s ability to perceive object position, scale, and occlusion. The experimental results show that ED-DETR achieves the current best 33.6% mAP on the AI-TOD-V2 dataset, which predominantly contains tiny objects, outperforming previous CNN-based and DETR-based methods, and shows excellent generalization performance on the VisDrone and COCO datasets.
- 14 January 2026
- Published by Elsevier in Neural Networks
- Vol. 198, 108597
A novel movie scene detection method based on clue relationship and constrained shot description
- 14 January 2026
- Published by Elsevier in Neural Networks
- Vol. 198, 108597
A novel movie scene detection method based on clue relationship and constrained shot description
- 8 September 2025
- Published by Springer Nature
- p. 71-114
Autonomous Vehicle Group: Formation, Evolution, and Cooperation in Internet of Vehicles
- 8 September 2025
- Published by Springer Nature
- p. 71-114
Autonomous Vehicle Group: Formation, Evolution, and Cooperation in Internet of Vehicles
- 30 July 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Consumer Electronics
- Vol. 71  (3) , 8617-8625
A Network Resource Slicing Method for Consumer Electronics Based on Holographic Counterpart Real Time Monitoring
The rapid development of Internet of things leads to a significant increase in connected consumer electronic counts, resulting in growing demands for network resources, e.g., bandwidth and spectrum. Their traditional allocation methods struggle in dynamic environments, exhibiting slow decision-making, low efficiency, poor slice isolation, and limited fine-grained slicing capabilities. To solve these issues, this paper proposes a network resource slicing method with holographic counterpart real time monitoring. First, it establishes a three-tier cellular heterogeneous network architecture based on holographic counterparts suitable for consumer electronics. Next, it constructs a mapping relationship between holographic counterparts and resource slices, enabling fine-grained resource allocation within this framework. Then, it introduces a solving method by integrating long short-term memory networks for time-series data processing and long-term dependency capture, along with the Advantage Actor-Critic (A2C) algorithm for decision optimization to address dynamic and complex nature of network environments. Experiments result show that our method outperforms traditional ones in spectral and energy efficiency and system utility. Specifically, it achieves 8.2 bps/W in energy efficiency, which is 3.9% and 6.5% higher than A2C and deep Q-network respectively.
- 30 July 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Consumer Electronics
- Vol. 71  (3) , 8617-8625
A Network Resource Slicing Method for Consumer Electronics Based on Holographic Counterpart Real Time Monitoring
The rapid development of Internet of things leads to a significant increase in connected consumer electronic counts, resulting in growing demands for network resources, e.g., bandwidth and spectrum. Their traditional allocation methods struggle in dynamic environments, exhibiting slow decision-making, low efficiency, poor slice isolation, and limited fine-grained slicing capabilities. To solve these issues, this paper proposes a network resource slicing method with holographic counterpart real time monitoring. First, it establishes a three-tier cellular heterogeneous network architecture based on holographic counterparts suitable for consumer electronics. Next, it constructs a mapping relationship between holographic counterparts and resource slices, enabling fine-grained resource allocation within this framework. Then, it introduces a solving method by integrating long short-term memory networks for time-series data processing and long-term dependency capture, along with the Advantage Actor-Critic (A2C) algorithm for decision optimization to address dynamic and complex nature of network environments. Experiments result show that our method outperforms traditional ones in spectral and energy efficiency and system utility. Specifically, it achieves 8.2 bps/W in energy efficiency, which is 3.9% and 6.5% higher than A2C and deep Q-network respectively.
- 18 July 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Mobile Computing
- Vol. 24  (12) , 13441-13458
Contributed Perception-Based Dynamic Evolution Method for Autonomous Vehicle Groups in Open Scenes
Accurately handling dynamic evolution events is a significant challenge for autonomous vehicle groups (AVGs) in open scenes, which can be affected by complex road conditions and various interference factors. Existing work on the dynamic evolution of AVGs in open scenes concentrates on semi-centralized groups, assessing communication links as the sole criterion. However, there lack the mathematical analysis of and methods for the dynamic evolution of events in distributed AVGs with cooperative perception. To address this issue, we propose a contributed perception-based dynamic evolution method designed for distributed AVGs. This method ensures that group members can continuously and timely exchange valid perceptual information. First, we investigate the impact of external interference on the contributed perception of vehicle groups to understand the drivers behind their dynamic evolution. Second, we define a range of vehicle group evolution behaviors and corresponding handling methods in response to external interference. Lastly, we introduce group states and perceptibility to delineate the evolution dynamics. Simulation results demonstrate the superiority of our proposed method over existing ones in terms of average group contribution, accessibility, persistence, timeliness, and perceptibility.
- 18 July 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Mobile Computing
- Vol. 24  (12) , 13441-13458
Contributed Perception-Based Dynamic Evolution Method for Autonomous Vehicle Groups in Open Scenes
Accurately handling dynamic evolution events is a significant challenge for autonomous vehicle groups (AVGs) in open scenes, which can be affected by complex road conditions and various interference factors. Existing work on the dynamic evolution of AVGs in open scenes concentrates on semi-centralized groups, assessing communication links as the sole criterion. However, there lack the mathematical analysis of and methods for the dynamic evolution of events in distributed AVGs with cooperative perception. To address this issue, we propose a contributed perception-based dynamic evolution method designed for distributed AVGs. This method ensures that group members can continuously and timely exchange valid perceptual information. First, we investigate the impact of external interference on the contributed perception of vehicle groups to understand the drivers behind their dynamic evolution. Second, we define a range of vehicle group evolution behaviors and corresponding handling methods in response to external interference. Lastly, we introduce group states and perceptibility to delineate the evolution dynamics. Simulation results demonstrate the superiority of our proposed method over existing ones in terms of average group contribution, accessibility, persistence, timeliness, and perceptibility.
- 26 June 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Consumer Electronics
- Vol. 71  (3) , 8571-8581
An Autonomous Vehicle Collaborative Perception Method Based on Holographic Counterpart Construction From Consumer Electronics Sensors
Accurate environmental perception is essential for autonomous vehicles. However, traditional sensors face limitations like obscured vision and blind spots. With the pursuit of high-precision sensors (such as cameras, radars, and lidars) in the fields of consumer electronics and holographic technology, more efficient perception devices are provided for autonomous vehicles. Therefore, this paper proposes a collaborative perception method for autonomous vehicles using consumer electronics sensors, which has three core modules: object detection, holographic counterpart construction, and collaborative perception. The object detection module enhances YOLO with a feature focus component, improving small object detection accuracy. The holographic counterpart module restores depth information using the pinhole camera principle, eliminating the need for dedicated depth sensors. In the collaborative perception module, vehicles share and fuse perception results with neighbors to form a holistic view of the environment. Experiments on three datasets show that this method outperforms five existing methods, achieving a 5.6% improvement in mAP@0.7 on OPV2V. The results highlight the potential of consumer electronics sensors and holographic perception to enhance autonomous vehicle safety and reliability.
- 26 June 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Consumer Electronics
- Vol. 71  (3) , 8571-8581
An Autonomous Vehicle Collaborative Perception Method Based on Holographic Counterpart Construction From Consumer Electronics Sensors
Accurate environmental perception is essential for autonomous vehicles. However, traditional sensors face limitations like obscured vision and blind spots. With the pursuit of high-precision sensors (such as cameras, radars, and lidars) in the fields of consumer electronics and holographic technology, more efficient perception devices are provided for autonomous vehicles. Therefore, this paper proposes a collaborative perception method for autonomous vehicles using consumer electronics sensors, which has three core modules: object detection, holographic counterpart construction, and collaborative perception. The object detection module enhances YOLO with a feature focus component, improving small object detection accuracy. The holographic counterpart module restores depth information using the pinhole camera principle, eliminating the need for dedicated depth sensors. In the collaborative perception module, vehicles share and fuse perception results with neighbors to form a holistic view of the environment. Experiments on three datasets show that this method outperforms five existing methods, achieving a 5.6% improvement in mAP@0.7 on OPV2V. The results highlight the potential of consumer electronics sensors and holographic perception to enhance autonomous vehicle safety and reliability.
- 9 April 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. 12  (13) , 25576-25586
Privacy-Preserving Autonomous Vehicle Group Formation in a Collusive Attack Scenario
The dynamic topologies and sensitive information exchanged among autonomous vehicle groups make them prime targets for attackers. In particular, in a collusive attack scenario, malicious nodes can collaborate to manipulate the trust evaluation system, thereby compromising the security of the entire vehicle group. To handle this limitation, this work proposes a privacy-preserving method for forming autonomous vehicle groups in a collusive attack scenario. First, we introduce a distributed trust evaluation algorithm based on a federated learning topology, which preserves local data privacy while facilitating reliable intervehicle trust computation. Then, we propose a PageRank-based detection mechanism that analyzes the trust propagation network to identify potential collusive attackers. Finally, we present a privacy-preserving method for autonomous vehicle group formation. Experimental results show that our proposed approach significantly improves the security and stability of autonomous vehicle groups compared to existing methods.
- 9 April 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. 12  (13) , 25576-25586
Privacy-Preserving Autonomous Vehicle Group Formation in a Collusive Attack Scenario
The dynamic topologies and sensitive information exchanged among autonomous vehicle groups make them prime targets for attackers. In particular, in a collusive attack scenario, malicious nodes can collaborate to manipulate the trust evaluation system, thereby compromising the security of the entire vehicle group. To handle this limitation, this work proposes a privacy-preserving method for forming autonomous vehicle groups in a collusive attack scenario. First, we introduce a distributed trust evaluation algorithm based on a federated learning topology, which preserves local data privacy while facilitating reliable intervehicle trust computation. Then, we propose a PageRank-based detection mechanism that analyzes the trust propagation network to identify potential collusive attackers. Finally, we present a privacy-preserving method for autonomous vehicle group formation. Experimental results show that our proposed approach significantly improves the security and stability of autonomous vehicle groups compared to existing methods.
- 21 March 2025
- Published by Springer Nature in Applied Intelligence
- Vol. 55  (7) , 1-22
HRMG-EA: Heterogeneous graph neural network recommendation with multi-level guidance based on enhanced-attributes
- 21 March 2025
- Published by Springer Nature in Applied Intelligence
- Vol. 55  (7) , 1-22
HRMG-EA: Heterogeneous graph neural network recommendation with multi-level guidance based on enhanced-attributes
- 30 August 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Intelligent Transportation Systems
- Vol. 24  (12) , 13852-13862
An Autonomous Vehicle Group Cooperation Model in an Urban Scene
Formulating a cooperative autonomous vehicle group is challenging in an urban scene that has complex road networks and diverse disturbance. Existing methods of vehicle cluster cooperation in a vehicular ad-hoc network cannot be applied to autonomous vehicles because the latter have different requirements for a vehicle group structure and communication quality. Existing studies focus on autonomous vehicle group cooperation in closed and highway scenes only. Their outcomes cannot be directly applied to an urban scene because of its complex road conditions, incomplete cooperation properties, and lack of a vehicle group size control strategy. In this work, we formulate a cooperation model for autonomous vehicle groups in such scene. First, we analyze cooperation criteria based on the non-colliding aggregate motion of flocks and deduce the connectivity, coupling, timeliness, evolvability, and adaptivity of a vehicle group, based on which we propose a cooperation model. Next, we solve our model by using a modified distributed evolutionary multi-objective optimization method, prove its convergence, and analyze its computational complexity. Finally, we conduct simulations on synthetic and real roads to show its performance in terms of average connectivity, coupling, timeliness, evolvability, and adaptivity of vehicle groups.
- 30 August 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Intelligent Transportation Systems
- Vol. 24  (12) , 13852-13862
An Autonomous Vehicle Group Cooperation Model in an Urban Scene
Formulating a cooperative autonomous vehicle group is challenging in an urban scene that has complex road networks and diverse disturbance. Existing methods of vehicle cluster cooperation in a vehicular ad-hoc network cannot be applied to autonomous vehicles because the latter have different requirements for a vehicle group structure and communication quality. Existing studies focus on autonomous vehicle group cooperation in closed and highway scenes only. Their outcomes cannot be directly applied to an urban scene because of its complex road conditions, incomplete cooperation properties, and lack of a vehicle group size control strategy. In this work, we formulate a cooperation model for autonomous vehicle groups in such scene. First, we analyze cooperation criteria based on the non-colliding aggregate motion of flocks and deduce the connectivity, coupling, timeliness, evolvability, and adaptivity of a vehicle group, based on which we propose a cooperation model. Next, we solve our model by using a modified distributed evolutionary multi-objective optimization method, prove its convergence, and analyze its computational complexity. Finally, we conduct simulations on synthetic and real roads to show its performance in terms of average connectivity, coupling, timeliness, evolvability, and adaptivity of vehicle groups.