Chao Xiang
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
Chao Xiang has a total of 28 co-authors. Here are the top co-authors by number of shared publications:
Mingchuan Yang
3 shared publications- Chen Feng
China Telecom Beijing Research Institute, Beijing, China
2 shared publications Feng Wang
2 shared publicationsLi Zhang
2 shared publications- Xiaopo Xie
China Telecom Beijing Research Institute, Beijing, China
2 shared publications - Zhendong Niu
Department of Computer Science and Technology, Beijing Institute of Technology, Beijing, China
2 shared publications
Latest Publications
- 1 December 2025
- Published by Elsevier in Information Fusion
- Vol. 129
Refinement-Guided Critique Learning: A Framework for Training Critique Models
- 1 December 2025
- Published by Elsevier in Information Fusion
- Vol. 129
Refinement-Guided Critique Learning: A Framework for Training Critique Models
- 22 May 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Intelligent Transportation Systems Magazine
- Vol. 17 (6) , 55-66
MTRCP: Multimodal Two-Level Fusion Architecture for Roadside Cooperative Perception
With the rapid development of self-driving technologies, autonomous driving still faces challenges because of the complexity of traffic environments, making accurate and stable environmental perception crucial. Roadside units (RSUs) can significantly enhance the perception range of autonomous vehicles, effectively addressing blind spots and improving traffic safety. This article proposes the Multimodal Two-Level Fusion Architecture for Roadside Cooperative Perception, a novel framework that leverages RSUs to enable comprehensive and over-the-horizon perception capabilities. We fuse point clouds and images at the first level to generate local detection results and then perform a second-level fusion to integrate these local results, producing full-area perception. Experimental validation with the DAIR-V2X-seq dataset and data collected at the China Telecom Smart Grid Test Park demonstrates the effectiveness and feasibility of the proposed cooperative perception architecture.
- 22 May 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Intelligent Transportation Systems Magazine
- Vol. 17 (6) , 55-66
MTRCP: Multimodal Two-Level Fusion Architecture for Roadside Cooperative Perception
With the rapid development of self-driving technologies, autonomous driving still faces challenges because of the complexity of traffic environments, making accurate and stable environmental perception crucial. Roadside units (RSUs) can significantly enhance the perception range of autonomous vehicles, effectively addressing blind spots and improving traffic safety. This article proposes the Multimodal Two-Level Fusion Architecture for Roadside Cooperative Perception, a novel framework that leverages RSUs to enable comprehensive and over-the-horizon perception capabilities. We fuse point clouds and images at the first level to generate local detection results and then perform a second-level fusion to integrate these local results, producing full-area perception. Experimental validation with the DAIR-V2X-seq dataset and data collected at the China Telecom Smart Grid Test Park demonstrates the effectiveness and feasibility of the proposed cooperative perception architecture.
- Top Cited Papers
- 4 August 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Intelligent Transportation Systems Magazine
- Vol. 15 (5) , 36-58
Multi-Sensor Fusion and Cooperative Perception for Autonomous Driving: A Review
Autonomous driving (AD), including single-vehicle intelligent AD and vehicle–infrastructure cooperative AD, has become a current research hot spot in academia and industry, and multi-sensor fusion is a fundamental task for AD system perception. However, the multi-sensor fusion process faces the problem of differences in the type and dimensionality of sensory data acquired using different sensors (cameras, lidar, millimeter-wave radar, and so on) as well as differences in the performance of environmental perception caused by using different fusion strategies. In this article, we study multiple papers on multi-sensor fusion in the field of AD and address the problem that the category division in current multi-sensor fusion perception is not detailed and clear enough and is more subjective, which makes the classification strategies differ significantly among similar algorithms. We innovatively propose a multi-sensor fusion taxonomy, which divides the fusion perception classification strategies into two categories—symmetric fusion and asymmetric fusion—and seven subcategories of strategy combinations, such as data, features, and results. In addition, the reliability of current AD perception is limited by its insufficient environment perception capability and the robustness of data-driven methods in dealing with extreme situations (e.g., blind areas). This article also summarizes the innovative applications of multi-sensor fusion classification strategies in AD cooperative perception.
- 4 August 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Intelligent Transportation Systems Magazine
- Vol. 15 (5) , 36-58
Top Cited PapersMulti-Sensor Fusion and Cooperative Perception for Autonomous Driving: A Review
Autonomous driving (AD), including single-vehicle intelligent AD and vehicle–infrastructure cooperative AD, has become a current research hot spot in academia and industry, and multi-sensor fusion is a fundamental task for AD system perception. However, the multi-sensor fusion process faces the problem of differences in the type and dimensionality of sensory data acquired using different sensors (cameras, lidar, millimeter-wave radar, and so on) as well as differences in the performance of environmental perception caused by using different fusion strategies. In this article, we study multiple papers on multi-sensor fusion in the field of AD and address the problem that the category division in current multi-sensor fusion perception is not detailed and clear enough and is more subjective, which makes the classification strategies differ significantly among similar algorithms. We innovatively propose a multi-sensor fusion taxonomy, which divides the fusion perception classification strategies into two categories—symmetric fusion and asymmetric fusion—and seven subcategories of strategy combinations, such as data, features, and results. In addition, the reliability of current AD perception is limited by its insufficient environment perception capability and the robustness of data-driven methods in dealing with extreme situations (e.g., blind areas). This article also summarizes the innovative applications of multi-sensor fusion classification strategies in AD cooperative perception.
- 10 July 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Computational Social Systems
- Vol. 10 (6) , 3066-3076
HPV-RCNN: Hybrid Point–Voxel Two-Stage Network for LiDAR-Based 3-D Object Detection
The current two-stage detectors remarkably benefit from hybrid representation of points and 3-D voxels, but they have high time cost and leave room for improving the accuracy of small objects. On the contrary, 2-D voxel-based methods tend to have good efficiency and better performance for small objects. An intuitive idea of optimizing a two-stage algorithm is to use a 2-D voxel-based backbone. However, naive representation substitution cannot achieve optimal joint learning of each representation and may cause a decrease in accuracy. In this article, we propose hybrid point–voxel RCNN (HPV-RCNN), a novel point cloud detection network which combines the merits of points and 2-D voxels. First, we propose a multiattentive voxel feature encoding module (MAVFE) to exploit multilevel attention of multiscale voxels. We also present a partial fusion pyramid network (PFPN) to effectively integrate multiresolution features and generate high-quality proposals. Then, a multiscale region of interest (RoI)-grid pooling (MSRGP) module is proposed to adaptively abstract proposal-specific features from sampled keypoints in multiple receptive fields. In addition, a cascade attentive module (CAM) is adopted to achieve incrementally proposal refinement by subsequent multiple subnetworks. Our method reaches top performance among two-stage methods in Cyclist and Pedestrian categories on the KITTI dataset while achieving real-time inference speed. Extensive experiments on challenging roadside DAIR-V2X-I dataset also demonstrate that our method achieves superior detection performance.
- 10 July 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Computational Social Systems
- Vol. 10 (6) , 3066-3076
HPV-RCNN: Hybrid Point–Voxel Two-Stage Network for LiDAR-Based 3-D Object Detection
The current two-stage detectors remarkably benefit from hybrid representation of points and 3-D voxels, but they have high time cost and leave room for improving the accuracy of small objects. On the contrary, 2-D voxel-based methods tend to have good efficiency and better performance for small objects. An intuitive idea of optimizing a two-stage algorithm is to use a 2-D voxel-based backbone. However, naive representation substitution cannot achieve optimal joint learning of each representation and may cause a decrease in accuracy. In this article, we propose hybrid point–voxel RCNN (HPV-RCNN), a novel point cloud detection network which combines the merits of points and 2-D voxels. First, we propose a multiattentive voxel feature encoding module (MAVFE) to exploit multilevel attention of multiscale voxels. We also present a partial fusion pyramid network (PFPN) to effectively integrate multiresolution features and generate high-quality proposals. Then, a multiscale region of interest (RoI)-grid pooling (MSRGP) module is proposed to adaptively abstract proposal-specific features from sampled keypoints in multiple receptive fields. In addition, a cascade attentive module (CAM) is adopted to achieve incrementally proposal refinement by subsequent multiple subnetworks. Our method reaches top performance among two-stage methods in Cyclist and Pedestrian categories on the KITTI dataset while achieving real-time inference speed. Extensive experiments on challenging roadside DAIR-V2X-I dataset also demonstrate that our method achieves superior detection performance.
- Open Access
- 30 May 2022
- Published by SAGE Publications in International Journal of Distributed Sensor Networks
- Vol. 18 (5)
Multi-sensor fusion algorithm in cooperative vehicle-infrastructure system for blind spot warning
With the rapid development of electric vehicles and artificial intelligence technology, the automatic driving industry has entered a rapid development stage. However, there is a risk of traffic accidents due to the blind spot of vision, whether autonomous vehicles or traditional vehicles. In this article, a multi-sensor fusion perception method is proposed, in which the semantic information from the camera and the range information from the LiDAR are fused at the data layer and the LiDAR point cloud containing semantic information is clustered to obtain the type and location information of the objects. Based on the sensor equipments deployed on the roadside, the sensing information processed by the fusion method is sent to the nearby vehicles in real-time through 5G and V2X technology for blind spot early warning, and its feasibility is verified by experiments and simulations. The blind spot warning scheme based on roadside multi-sensor fusion perception proposed in this article has been experimentally verified in the closed park, which can obviously reduce the traffic accidents caused by the blind spot of vision, and is of great significance to improve traffic safety.
- 30 May 2022
- Published by SAGE Publications in International Journal of Distributed Sensor Networks
- Vol. 18 (5)
Open AccessMulti-sensor fusion algorithm in cooperative vehicle-infrastructure system for blind spot warning
With the rapid development of electric vehicles and artificial intelligence technology, the automatic driving industry has entered a rapid development stage. However, there is a risk of traffic accidents due to the blind spot of vision, whether autonomous vehicles or traditional vehicles. In this article, a multi-sensor fusion perception method is proposed, in which the semantic information from the camera and the range information from the LiDAR are fused at the data layer and the LiDAR point cloud containing semantic information is clustered to obtain the type and location information of the objects. Based on the sensor equipments deployed on the roadside, the sensing information processed by the fusion method is sent to the nearby vehicles in real-time through 5G and V2X technology for blind spot early warning, and its feasibility is verified by experiments and simulations. The blind spot warning scheme based on roadside multi-sensor fusion perception proposed in this article has been experimentally verified in the closed park, which can obviously reduce the traffic accidents caused by the blind spot of vision, and is of great significance to improve traffic safety.
- 2 November 2016
- Published by Springer Nature in Landscape and Ecological Engineering
- Vol. 13 (1) , 169-181
Evaluating the effectiveness of landscape configuration metrics from landscape composition metrics
- 2 November 2016
- Published by Springer Nature in Landscape and Ecological Engineering
- Vol. 13 (1) , 169-181
Evaluating the effectiveness of landscape configuration metrics from landscape composition metrics