Qiyan Li
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Qiyan Li has a total of 14 co-authors. Here are the top co-authors by number of shared publications:
- Chang Wu
Institute of Information and Communication Engineering, University of Electronic Science and Technology of China (UESTC), Chengdu, China
3 shared publications - Xiaotong Kong
Institute of Information and Communication Engineering, University of Electronic Science and Technology of China (UESTC), Chengdu, China
2 shared publications - Junhai Luo
Laboratory of Imaging Detection and Intelligent Perception and the School of Information and Communication Engineering, University of Electronic Science and Technology of China, Chengdu, China
1 shared publication - Kaiyong Zhao
XGRIDS, Shenzhen, China
1 shared publication - Lang Ming
Institute of Information and Communication Engineering, University of Electronic Science and Technology of China (UESTC), Chengdu, China
1 shared publication - Qiucen Li
Institute of Information and Communication Engineering, University of Electronic Science and Technology of China (UESTC), Chengdu, China
1 shared publication
Latest Publications
- 24 March 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. 13 (11) , 25016-25033
VID-SLAM: A Robust Visual–Inertial–DVL Tightly Coupled Localization Method for Underwater Robots
Simultaneous localization and mapping (SLAM) has emerged as a promising solution to address the localization challenges faced by underwater robots. This work proposes a factor graph optimization-based visual–inertial–Doppler velocity log (DVL) tightly coupled SLAM method designed for underwater robot localization. The approach introduces a novel DVL residual construction method to maximize the utility of DVL measurements. IMU data is integrated to detect anomalies in DVL measurements, and a DVL synchronization marking strategy is developed to enhance the method’s applicability and strengthen data associations. To improve system robustness in scenarios where visual tracking fails, a new sliding window strategy is incorporated. Experimental results on underwater datasets and simulations demonstrate that the proposed method achieves significant improvements in localization accuracy and robustness compared to existing underwater localization algorithms. The implementation code of the proposed method and a simulated underwater dataset can be accessed at https://github.com/uestc-icsp/VID-SLAM
- 24 March 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. 13 (11) , 25016-25033
VID-SLAM: A Robust Visual–Inertial–DVL Tightly Coupled Localization Method for Underwater Robots
Simultaneous localization and mapping (SLAM) has emerged as a promising solution to address the localization challenges faced by underwater robots. This work proposes a factor graph optimization-based visual–inertial–Doppler velocity log (DVL) tightly coupled SLAM method designed for underwater robot localization. The approach introduces a novel DVL residual construction method to maximize the utility of DVL measurements. IMU data is integrated to detect anomalies in DVL measurements, and a DVL synchronization marking strategy is developed to enhance the method’s applicability and strengthen data associations. To improve system robustness in scenarios where visual tracking fails, a new sliding window strategy is incorporated. Experimental results on underwater datasets and simulations demonstrate that the proposed method achieves significant improvements in localization accuracy and robustness compared to existing underwater localization algorithms. The implementation code of the proposed method and a simulated underwater dataset can be accessed at https://github.com/uestc-icsp/VID-SLAM
- 16 May 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Intelligent Transportation Systems
- Vol. 26 (7) , 9918-9931
Light-SLAM: A Robust Deep-Learning Visual SLAM System Based on LightGlue Under Challenging Lighting Conditions
Simultaneous Localization and Mapping (SLAM) has become a critical technology for intelligent transportation systems and autonomous robots and is widely used in autonomous driving. However, traditional manual feature-based methods in challenging lighting environments make it difficult to ensure robustness and accuracy. Some deep learning-based methods show potential but still have significant drawbacks. To address this problem, we propose a feature-based visual SLAM system based on the LightGlue deep learning network. It uses deep local feature descriptors to replace traditional hand-crafted features and a more efficient and accurate deep network to achieve fast and precise feature matching. Thus, we use the robustness of deep learning to improve the whole system. We have combined traditional geometry-based approaches to introduce a complete visual SLAM system for monocular, binocular, and RGB-D sensors. We thoroughly tested the proposed system on four public datasets: KITTI, EuRoC, TUM, and 4Season, as well as on actual campus scenes. The experimental results show that the proposed method exhibits better accuracy and robustness in adapting to low-light and strongly light-varying environments than traditional manual features and deep learning-based methods. It can also run on GPU in real time.
- 16 May 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Intelligent Transportation Systems
- Vol. 26 (7) , 9918-9931
Light-SLAM: A Robust Deep-Learning Visual SLAM System Based on LightGlue Under Challenging Lighting Conditions
Simultaneous Localization and Mapping (SLAM) has become a critical technology for intelligent transportation systems and autonomous robots and is widely used in autonomous driving. However, traditional manual feature-based methods in challenging lighting environments make it difficult to ensure robustness and accuracy. Some deep learning-based methods show potential but still have significant drawbacks. To address this problem, we propose a feature-based visual SLAM system based on the LightGlue deep learning network. It uses deep local feature descriptors to replace traditional hand-crafted features and a more efficient and accurate deep network to achieve fast and precise feature matching. Thus, we use the robustness of deep learning to improve the whole system. We have combined traditional geometry-based approaches to introduce a complete visual SLAM system for monocular, binocular, and RGB-D sensors. We thoroughly tested the proposed system on four public datasets: KITTI, EuRoC, TUM, and 4Season, as well as on actual campus scenes. The experimental results show that the proposed method exhibits better accuracy and robustness in adapting to low-light and strongly light-varying environments than traditional manual features and deep learning-based methods. It can also run on GPU in real time.
- 17 November 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Robotics and Automation Letters
- Vol. 9 (1) , 427-434
VoxelMap++: Mergeable Voxel Mapping Method for Online LiDAR(-Inertial) Odometry
This letter presents VoxelMap++: a voxel mapping method with plane merging which can effectively improve the accuracy and efficiency of LiDAR(-inertial) based simultaneous localization and mapping (SLAM). This map is a collection of voxels that contains one plane feature with 3DOF representation and corresponding covariance estimation. Considering map will contain a large number of coplanar features (kid planes), these kid planes' can be regarded as the measurements with covariance of a larger plane (father plane). Thus, we have designed a plane merging module based on the union-find. This merging module is capable of distinguishing co-plane relationship within various voxels, then merge these kid planes to estimate the father plane by minimizing the trace of covariance. After merging, the father plane exhibits more accurate compare to kids plane, with decreasing of uncertainty, which improve the accuracy of LiDAR(-inertial) odometry. Experiments on different environments demonstrate the superior of VoxelMap++ compared with other state-of-the-art methods (see our attached video). Our implementation is open-sourced on GitHub which is applicable for both non-repetitive scanning LiDARs and traditional scanning LiDAR.
- 17 November 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Robotics and Automation Letters
- Vol. 9 (1) , 427-434
VoxelMap++: Mergeable Voxel Mapping Method for Online LiDAR(-Inertial) Odometry
This letter presents VoxelMap++: a voxel mapping method with plane merging which can effectively improve the accuracy and efficiency of LiDAR(-inertial) based simultaneous localization and mapping (SLAM). This map is a collection of voxels that contains one plane feature with 3DOF representation and corresponding covariance estimation. Considering map will contain a large number of coplanar features (kid planes), these kid planes' can be regarded as the measurements with covariance of a larger plane (father plane). Thus, we have designed a plane merging module based on the union-find. This merging module is capable of distinguishing co-plane relationship within various voxels, then merge these kid planes to estimate the father plane by minimizing the trace of covariance. After merging, the father plane exhibits more accurate compare to kids plane, with decreasing of uncertainty, which improve the accuracy of LiDAR(-inertial) odometry. Experiments on different environments demonstrate the superior of VoxelMap++ compared with other state-of-the-art methods (see our attached video). Our implementation is open-sourced on GitHub which is applicable for both non-repetitive scanning LiDARs and traditional scanning LiDAR.