Revisiting visual odometry for real-time performance
Visual Odometry (VO) is a key component in modern driver assistance systems and robotics. Meeting the real-time requirements is mandatory for VO in such applications. Previous works have primarily focused on improving accuracy at the cost of longer runtime. In this work, we propose novel strategies...
| Main Authors: | , , |
|---|---|
| Other Authors: | |
| Format: | Conference Paper |
| Language: | English |
| Published: |
2024
|
| Subjects: | |
| Online Access: | https://hdl.handle.net/10356/178589 |
| _version_ | 1824453116088549376 |
|---|---|
| author | Singh, Gaurav Wu, Meiqing Lam, Siew-Kei |
| author2 | College of Computing and Data Science |
| author_facet | College of Computing and Data Science Singh, Gaurav Wu, Meiqing Lam, Siew-Kei |
| author_sort | Singh, Gaurav |
| collection | NTU |
| description | Visual Odometry (VO) is a key component in modern driver assistance systems and robotics. Meeting the real-time requirements is mandatory for VO in such applications. Previous works have primarily focused on improving accuracy at the cost of longer runtime. In this work, we propose novel strategies for feature correspondence setup, outlier removal and robust pose optimization in the VO pipeline to achieve real-time performance of close to 30 frames-per-seconds (fps) on a dual-core 3.5 GHz CPU while maintaining high accuracy. In particular, computationally efficient strategies are introduced to obtain an initial set of good features and rapidly filter out the outliers to minimize the computational overhead in later stages. In addition, we propose a depth based weighting and saturated-residual scheme during pose optimization to increase the robustness of VO. Experimental results show that the proposed VO achieves the fastest speed among all the top-ranked OV and SLAM systems on KITTI leader-board. Specifically, the proposed VO is 47% faster than state-of-the-art ORB-SLAM2 with comparable accuracy on KITTI dataset. |
| first_indexed | 2024-10-01T07:55:17Z |
| format | Conference Paper |
| id | ntu-10356/178589 |
| institution | Nanyang Technological University |
| language | English |
| last_indexed | 2024-10-01T07:55:17Z |
| publishDate | 2024 |
| record_format | dspace |
| spelling | ntu-10356/1785892024-06-27T01:35:25Z Revisiting visual odometry for real-time performance Singh, Gaurav Wu, Meiqing Lam, Siew-Kei College of Computing and Data Science School of Computer Science and Engineering 2019 16th International Conference on Machine Vision Applications (MVA) Computer and Information Science Feature extraction Optimization Visual Odometry (VO) is a key component in modern driver assistance systems and robotics. Meeting the real-time requirements is mandatory for VO in such applications. Previous works have primarily focused on improving accuracy at the cost of longer runtime. In this work, we propose novel strategies for feature correspondence setup, outlier removal and robust pose optimization in the VO pipeline to achieve real-time performance of close to 30 frames-per-seconds (fps) on a dual-core 3.5 GHz CPU while maintaining high accuracy. In particular, computationally efficient strategies are introduced to obtain an initial set of good features and rapidly filter out the outliers to minimize the computational overhead in later stages. In addition, we propose a depth based weighting and saturated-residual scheme during pose optimization to increase the robustness of VO. Experimental results show that the proposed VO achieves the fastest speed among all the top-ranked OV and SLAM systems on KITTI leader-board. Specifically, the proposed VO is 47% faster than state-of-the-art ORB-SLAM2 with comparable accuracy on KITTI dataset. 2024-06-27T01:35:25Z 2024-06-27T01:35:25Z 2019 Conference Paper Singh, G., Wu, M. & Lam, S. (2019). Revisiting visual odometry for real-time performance. 2019 16th International Conference on Machine Vision Applications (MVA). https://dx.doi.org/10.23919/MVA.2019.8757936 978-4-901122-18-4 https://hdl.handle.net/10356/178589 10.23919/MVA.2019.8757936 2-s2.0-85070442237 en © 2019 MVA Organization. Published by IEEE. All rights reserved. |
| spellingShingle | Computer and Information Science Feature extraction Optimization Singh, Gaurav Wu, Meiqing Lam, Siew-Kei Revisiting visual odometry for real-time performance |
| title | Revisiting visual odometry for real-time performance |
| title_full | Revisiting visual odometry for real-time performance |
| title_fullStr | Revisiting visual odometry for real-time performance |
| title_full_unstemmed | Revisiting visual odometry for real-time performance |
| title_short | Revisiting visual odometry for real-time performance |
| title_sort | revisiting visual odometry for real time performance |
| topic | Computer and Information Science Feature extraction Optimization |
| url | https://hdl.handle.net/10356/178589 |
| work_keys_str_mv | AT singhgaurav revisitingvisualodometryforrealtimeperformance AT wumeiqing revisitingvisualodometryforrealtimeperformance AT lamsiewkei revisitingvisualodometryforrealtimeperformance |