Abstract | ||
---|---|---|
Autonomous driving in underground garages usually utilizes a 2D/3D occupancy map for localization. However, the real scene is changing, and may not be consistent with the map. Vehicles and other objects not contained in the map are considered as obstacles, which increase the difficulty of localization and affect the accuracy of result. In this paper, we propose a cylinder rotational projection statistics (Cy-RoPS) feature descriptor, which is a local surface feature descriptor to improve the accuracy of localization. The local surface feature motivated by RoPS feature is invariant to rotation of point set enclosed in a cylinder. We also propose to employ the local surface feature for localization in a real underground garage. The experimental results show that the proposed method is robust to dynamic obstacles in the underground garage, and has a higher accuracy in localization, compared with the state-of-the-art methods. |
Year | DOI | Venue |
---|---|---|
2018 | 10.1109/IVS.2018.8500492 | 2018 IEEE Intelligent Vehicles Symposium (IV) |
Keywords | Field | DocType |
underground garage,map matching,cylinder rotational projection statistics,local surface feature descriptor,RoPS feature | Computer vision,Computer science,Cylinder,Feature extraction,Lidar,Global Positioning System,Invariant (mathematics),Artificial intelligence,Feature based,Map matching,Cognitive neuroscience of visual object recognition | Conference |
ISSN | ISBN | Citations |
1931-0587 | 978-1-5386-4453-9 | 0 |
PageRank | References | Authors |
0.34 | 12 | 6 |
Name | Order | Citations | PageRank |
---|---|---|---|
Zhongxing Tao | 1 | 0 | 0.68 |
Jianru Xu | 2 | 0 | 0.34 |
Di Wang | 3 | 1337 | 143.48 |
Shuyang Zhang | 4 | 13 | 1.97 |
Dixiao Cui | 5 | 18 | 2.41 |
Shaoyi Du | 6 | 357 | 40.68 |