Chun Ouyang
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
Chun Ouyang has a total of 49 co-authors. Here are the top co-authors by number of shared publications:
- Zhongxue Gan
Fudan University
9 shared publications Linqiang Hu
7 shared publicationsYuning Chen
7 shared publicationsZhongxue Gan
7 shared publications- Hongda Zhang
Academy for Engineering and Technology, Fudan University, Shanghai, China
5 shared publications Wei Lin
Academy for Engineering and Technology, Fudan University, Shanghai, China
4 shared publications
Latest Publications
- Open Access
- 9 November 2026
- Published by Association for Computing Machinery (ACM)
- p. 7690-7698
Mitigating Shared-Private Branch Imbalance via Dual-Branch Rebalancing for Multimodal Sentiment Analysis
- 9 November 2026
- Published by Association for Computing Machinery (ACM)
- p. 7690-7698
Open AccessMitigating Shared-Private Branch Imbalance via Dual-Branch Rebalancing for Multimodal Sentiment Analysis
- Open Access
- 15 June 2026
- Published by Association for Computing Machinery (ACM)
- p. 1573-1577
Joint-Guided Spatial and Semantic Sensitive Diffusion Policy for Robotic Manipulation
- 15 June 2026
- Published by Association for Computing Machinery (ACM)
- p. 1573-1577
Open AccessJoint-Guided Spatial and Semantic Sensitive Diffusion Policy for Robotic Manipulation
- Open Access
- 13 March 2026
- Published by Elsevier in SSRN Electronic Journal
Sparse Large-Scale Optimization for Adaptive Flocking Coverage
Multi-robot coverage has become a critical technology for applications such as autonomous inspection, search and rescue, and precision agriculture, where efficient and safe exploration of unknown environments is essential. Optimization-based approaches have shown promise in coordinating large-scale swarms by tuning control parameters to balance multiple objectives. However, existing methods impose substantial computational burdens to simultaneously maintain exploration efficiency and inter-robot collision avoidance. In large-scale swarms, control parameters often exhibit intrinsic interdependencies, whereby only a sparse subset of decision variables significantly influences system performance. Leveraging this sparsity structure to reduce computational complexity while preserving coordination quality remains largely unexplored. To address this challenge, we propose the Sparse Large-Scale Optimized Flocking Coverage Framework (SLOFCF). First, we develop a distributed flocking coverage model with context-aware adaptive weight tuning that enables each robot to dynamically modulate interaction strengths based on local environmental information. The model integrates a lightweight frontier selection mechanism with time-to-collision prediction and cooperative hazard sharing, enabling effective coverage in dynamic environments. Second, the Sparse Large-Scale Evolutionary Algorithm with Mixed Importance Awareness (SLEMIA) identifies and optimizes only critical control variables through dominance-based ranking and perturbation-based sensitivity analysis, substantially reducing computational overhead. Extensive simulations demonstrate that SLOFCF achieves 20% average improvement in coverage efficiency while consistently outperforming baseline methods in task completion time, inter-robot safety, and obstacle avoidance.
- 13 March 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessSparse Large-Scale Optimization for Adaptive Flocking Coverage
Multi-robot coverage has become a critical technology for applications such as autonomous inspection, search and rescue, and precision agriculture, where efficient and safe exploration of unknown environments is essential. Optimization-based approaches have shown promise in coordinating large-scale swarms by tuning control parameters to balance multiple objectives. However, existing methods impose substantial computational burdens to simultaneously maintain exploration efficiency and inter-robot collision avoidance. In large-scale swarms, control parameters often exhibit intrinsic interdependencies, whereby only a sparse subset of decision variables significantly influences system performance. Leveraging this sparsity structure to reduce computational complexity while preserving coordination quality remains largely unexplored. To address this challenge, we propose the Sparse Large-Scale Optimized Flocking Coverage Framework (SLOFCF). First, we develop a distributed flocking coverage model with context-aware adaptive weight tuning that enables each robot to dynamically modulate interaction strengths based on local environmental information. The model integrates a lightweight frontier selection mechanism with time-to-collision prediction and cooperative hazard sharing, enabling effective coverage in dynamic environments. Second, the Sparse Large-Scale Evolutionary Algorithm with Mixed Importance Awareness (SLEMIA) identifies and optimizes only critical control variables through dominance-based ranking and perturbation-based sensitivity analysis, substantially reducing computational overhead. Extensive simulations demonstrate that SLOFCF achieves 20% average improvement in coverage efficiency while consistently outperforming baseline methods in task completion time, inter-robot safety, and obstacle avoidance.
- 1 March 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Biomedical and Health Informatics
- Vol. 30 (3) , 2537-2549
ViG3D-UNet: Volumetric Vascular Connectivity-Aware Segmentation via 3D Vision Graph Representation
Accurate vascular segmentation is essential for coronary visualization and the diagnosis of coronary heart disease. This task involves the extraction of sparse tree-like vascular branches from volumetric space. However, existing methods have faced significant challenges due to discontinuous vascular segmentation and missing endpoints. To address this issue, a 3D vision graph neural network framework, named ViG3D-UNet, was introduced. This method integrates 3D graph representation and aggregation within a U-shaped architecture to facilitate continuous vascular segmentation. The ViG3D module captures volumetric vascular connectivity and topology, while the convolutional module extracts fine vascular details. These two branches are combined through channel attention to form the encoder feature. Subsequently, a paperclip-shaped offset decoder minimizes redundant computations in the sparse feature space and restores the feature map size to match the original input dimensions. To evaluate the effectiveness of the proposed approach for continuous vascular segmentation, evaluations were performed on two public datasets, ASOCA and ImageCAS. The segmentation results show that the ViG3D-UNet surpassed competing methods in maintaining vascular segmentation connectivity while achieving high segmentation accuracy.
- 1 March 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Biomedical and Health Informatics
- Vol. 30 (3) , 2537-2549
ViG3D-UNet: Volumetric Vascular Connectivity-Aware Segmentation via 3D Vision Graph Representation
Accurate vascular segmentation is essential for coronary visualization and the diagnosis of coronary heart disease. This task involves the extraction of sparse tree-like vascular branches from volumetric space. However, existing methods have faced significant challenges due to discontinuous vascular segmentation and missing endpoints. To address this issue, a 3D vision graph neural network framework, named ViG3D-UNet, was introduced. This method integrates 3D graph representation and aggregation within a U-shaped architecture to facilitate continuous vascular segmentation. The ViG3D module captures volumetric vascular connectivity and topology, while the convolutional module extracts fine vascular details. These two branches are combined through channel attention to form the encoder feature. Subsequently, a paperclip-shaped offset decoder minimizes redundant computations in the sparse feature space and restores the feature map size to match the original input dimensions. To evaluate the effectiveness of the proposed approach for continuous vascular segmentation, evaluations were performed on two public datasets, ASOCA and ImageCAS. The segmentation results show that the ViG3D-UNet surpassed competing methods in maintaining vascular segmentation connectivity while achieving high segmentation accuracy.
- 23 September 2025
- Published by Elsevier in Journal of Industrial Information Integration
- Vol. 48
Multi-grained teacher–student joint representation learning for surface defect classification
- 23 September 2025
- Published by Elsevier in Journal of Industrial Information Integration
- Vol. 48
Multi-grained teacher–student joint representation learning for surface defect classification
- 9 June 2025
- Published by Elsevier in Applied Soft Computing
- Vol. 180
Dynamic cooperative coevolution based on variable importance for non-separable large-scale global optimization
- 9 June 2025
- Published by Elsevier in Applied Soft Computing
- Vol. 180
Dynamic cooperative coevolution based on variable importance for non-separable large-scale global optimization
- 9 May 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Evolutionary Computation
- Vol. 30 (2) , 795-809
Dynamic Grouping With a Self-Aware Computational Resource Allocation for Large-Scale Multi-Objective Optimization
In real-world applications, many problems are characterized by numerous variables and conflicting objectives, which are referred to as Large-Scale Multi-Objective Optimization Problems (LSMOPs). The Cooperative Coevolution (CC) algorithm is commonly used to solve LSMOPs. The decision variables of the target problem are partitioned into groups with each group being solved independently. However, existing multi-objective CC algorithms encounter difficulties in accurately grouping variables by importance and lack effective methods for allocating computational resources to enhance solution efficiency. To address these challenges, a dynamic grouping method based on variable importance is introduced alongside a self-aware computational resource allocation mechanism. Our proposed algorithm calculates the importance of each variable by considering the impact of value fluctuations on each objective. Variables with similar importance levels are subsequently grouped together. Based on this, the resource allocation mechanism is proposed to monitor changes in the improvement rate of population quality during the optimization process across different variable groups, dynamically directing resources to groups of greater importance. The proposed algorithm was tested against five state-of-the-art large-scale multi-objective algorithms (RG-DRA, LERD, LMOCSO, LMEA, FDV) using the LSMOP and WFG benchmark suites. It outperformed at least 75% of LSMOP instances and surpassed RG-DRA, LERD, and LMOCSO in over 50% of WFG instances, demonstrating comparable performance to LMEA and FDV.
- 9 May 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Evolutionary Computation
- Vol. 30 (2) , 795-809
Dynamic Grouping With a Self-Aware Computational Resource Allocation for Large-Scale Multi-Objective Optimization
In real-world applications, many problems are characterized by numerous variables and conflicting objectives, which are referred to as Large-Scale Multi-Objective Optimization Problems (LSMOPs). The Cooperative Coevolution (CC) algorithm is commonly used to solve LSMOPs. The decision variables of the target problem are partitioned into groups with each group being solved independently. However, existing multi-objective CC algorithms encounter difficulties in accurately grouping variables by importance and lack effective methods for allocating computational resources to enhance solution efficiency. To address these challenges, a dynamic grouping method based on variable importance is introduced alongside a self-aware computational resource allocation mechanism. Our proposed algorithm calculates the importance of each variable by considering the impact of value fluctuations on each objective. Variables with similar importance levels are subsequently grouped together. Based on this, the resource allocation mechanism is proposed to monitor changes in the improvement rate of population quality during the optimization process across different variable groups, dynamically directing resources to groups of greater importance. The proposed algorithm was tested against five state-of-the-art large-scale multi-objective algorithms (RG-DRA, LERD, LMOCSO, LMEA, FDV) using the LSMOP and WFG benchmark suites. It outperformed at least 75% of LSMOP instances and surpassed RG-DRA, LERD, and LMOCSO in over 50% of WFG instances, demonstrating comparable performance to LMEA and FDV.
- 24 January 2025
- Published by Elsevier in Neurocomputing
- Vol. 624
RPN: A region-to-pixel-mask-based convolutional network for lesion segmentation of fundus images
- 24 January 2025
- Published by Elsevier in Neurocomputing
- Vol. 624
RPN: A region-to-pixel-mask-based convolutional network for lesion segmentation of fundus images
- 20 January 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Industrial Informatics
- Vol. 21 (4) , 3007-3016
Real-Time Scheduling Framework for Multiagent Cooperative Logistics With Dynamic Supply Demands
In logistics systems with multiagent collaboration, one of the prevailing focus lies on modeling as the dynamic multiperiod vehicle routing problem (DMPVRP). This work introduces modifications to DMPVRP to align with the requirements of real factory operations, particularly with dynamic supply demands. A self-established multiagent dynamic scheduling framework has been proposed to adapt to dynamic environmental changes and make timely adjustments, which consists of two modules: dynamic path planning and machine assignment. The first module utilizes a self-designed multioperator two-stage evolutionary algorithm to dynamically update the routes for vehicles. The second module maintains the workload balance among vehicles in real time. Experimental results demonstrate that the proposed algorithm achieves optimal outcomes compared to three state-of-the-art algorithms, surpassing others by 20% in machine output and exhibiting 5% lower transportation costs. In addition, a case study from a steel cord manufacturing factory is conducted, demonstrating its capability to promptly enhance efficiency.
- 20 January 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Industrial Informatics
- Vol. 21 (4) , 3007-3016
Real-Time Scheduling Framework for Multiagent Cooperative Logistics With Dynamic Supply Demands
In logistics systems with multiagent collaboration, one of the prevailing focus lies on modeling as the dynamic multiperiod vehicle routing problem (DMPVRP). This work introduces modifications to DMPVRP to align with the requirements of real factory operations, particularly with dynamic supply demands. A self-established multiagent dynamic scheduling framework has been proposed to adapt to dynamic environmental changes and make timely adjustments, which consists of two modules: dynamic path planning and machine assignment. The first module utilizes a self-designed multioperator two-stage evolutionary algorithm to dynamically update the routes for vehicles. The second module maintains the workload balance among vehicles in real time. Experimental results demonstrate that the proposed algorithm achieves optimal outcomes compared to three state-of-the-art algorithms, surpassing others by 20% in machine output and exhibiting 5% lower transportation costs. In addition, a case study from a steel cord manufacturing factory is conducted, demonstrating its capability to promptly enhance efficiency.
- 2 January 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Biomedical and Health Informatics
- Vol. 29 (5) , 3576-3586
RTS-ViT: Real-Time Share Vision Transformer for Image Classification
Vision transformers have achieved remarkable success in image classification. The dual-branch vision transformer generates more features by taking advantage of feature fusion. Inspired by this, a dual-branch vision transformer with Real-Time Share feature was proposed during the encoding process for retinal image classification tasks. The approach processes image patches of varying sizes (base and large) through two independent branches and implements multi-stage Real-Time feature fusion via the Real-Time Share feature encoder. This encoder enables the branches to complement each other's features at each encoding stage, facilitating finer feature learning and enhancing the self-attention information passed to subsequent stages. It significantly boosts feature representation and classification performance. Additionally, a straightforward and effective feature fusion method, L -Times Attention Fusion, was proposed: vector concatenation for Real-Time Share feature in the earlier ( L -1) encoding stages and element-wise addition for overall feature fusion at the L -th stage, achieving more efficient feature integration. The method was validated on a retinal image dataset. Results show that the approach outperforms the recent Cross-ViT average TOP-1 Acc by 5.61% with lower FLOPs and model parameters, without relying on pre-trained weights, highlighting stronger self-learning feature capabilities and reduced reliance on extensive pre-training data.
- 2 January 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Biomedical and Health Informatics
- Vol. 29 (5) , 3576-3586
RTS-ViT: Real-Time Share Vision Transformer for Image Classification
Vision transformers have achieved remarkable success in image classification. The dual-branch vision transformer generates more features by taking advantage of feature fusion. Inspired by this, a dual-branch vision transformer with Real-Time Share feature was proposed during the encoding process for retinal image classification tasks. The approach processes image patches of varying sizes (base and large) through two independent branches and implements multi-stage Real-Time feature fusion via the Real-Time Share feature encoder. This encoder enables the branches to complement each other's features at each encoding stage, facilitating finer feature learning and enhancing the self-attention information passed to subsequent stages. It significantly boosts feature representation and classification performance. Additionally, a straightforward and effective feature fusion method, L -Times Attention Fusion, was proposed: vector concatenation for Real-Time Share feature in the earlier ( L -1) encoding stages and element-wise addition for overall feature fusion at the L -th stage, achieving more efficient feature integration. The method was validated on a retinal image dataset. Results show that the approach outperforms the recent Cross-ViT average TOP-1 Acc by 5.61% with lower FLOPs and model parameters, without relying on pre-trained weights, highlighting stronger self-learning feature capabilities and reduced reliance on extensive pre-training data.