講演抄録/キーワード |
講演名 |
2016-12-02 10:05
3D Scene Understanding at Urban Intersections for Autonomous Vehicles Prarthana Bhattacharyya・○Yanlei Gu・Jiali Bao・Shunsuke Kamijo(UTokyo) WBS2016-69 ITS2016-36 RCC2016-58 |
抄録 |
(和) |
Urban intersections exhibit very complex driving behavior and are extremely prone to traffic accidents and collisions. It is thus very important for autonomous vehicles to comprehensively analyze and interpret the traffic suroundings before it can plan its subsequent motion. This paper presents a stereo vision and digital map based approach for autonomous vehicles to geometrically, contextually and semantically understand the 3D driving environment at urban intersections. Stereo vision is used to detect, classify and track dynamic obstacles, while digital 2D and 3D maps of the environment are used to identify static obstacles and provide context to the dynamic traffic participants. This paper also addresses the crucial aspect of ego-position and ego-motion estimation for accurate scene understanding systems by proposing to integrate the global navigation satellite systems (GNSS), inertial navigation sensors (INS) and on-board camera. Experiments conducted in various challenging conditions at Hitotsubashi, Tokyo demonstrate the effectiveness of the proposed system in providing comprehensive awareness of the traffic surroundings. |
(英) |
Urban intersections exhibit very complex driving behavior and are extremely prone to traffic accidents and collisions. It is thus very important for autonomous vehicles to comprehensively analyze and interpret the traffic suroundings before it can plan its subsequent motion. This paper presents a stereo vision and digital map based approach for autonomous vehicles to geometrically, contextually and semantically understand the 3D driving environment at urban intersections. Stereo vision is used to detect, classify and track dynamic obstacles, while digital 2D and 3D maps of the environment are used to identify static obstacles and provide context to the dynamic traffic participants. This paper also addresses the crucial aspect of ego-position and ego-motion estimation for accurate scene understanding systems by proposing to integrate the global navigation satellite systems (GNSS), inertial navigation sensors (INS) and on-board camera. Experiments conducted in various challenging conditions at Hitotsubashi, Tokyo demonstrate the effectiveness of the proposed system in providing comprehensive awareness of the traffic surroundings. |
キーワード |
(和) |
3D Scene Understanding / Digital Map / Stereo Vision / Integration / Vehicle Self-localization / / / |
(英) |
3D Scene Understanding / Digital Map / Stereo Vision / Integration / Vehicle Self-localization / / / |
文献情報 |
信学技報, vol. 116, no. 338, ITS2016-36, pp. 125-130, 2016年12月. |
資料番号 |
ITS2016-36 |
発行日 |
2016-11-24 (WBS, ITS, RCC) |
ISSN |
Print edition: ISSN 0913-5685 Online edition: ISSN 2432-6380 |
著作権に ついて |
技術研究報告に掲載された論文の著作権は電子情報通信学会に帰属します.(許諾番号:10GA0019/12GB0052/13GB0056/17GB0034/18GB0034) |
PDFダウンロード |
WBS2016-69 ITS2016-36 RCC2016-58 |
|