Variants for Visual Odometry Estimation

Title:Variants for Visual Odometry Estimation Date: 14:00-15:00, Sept 12 2018 Venue: N3-332 Qingdao

Speaker: Prof. Reinhard Klette
Date and Time: 14:00-15:00, Sept 12 2018
Venue:N3-332 Qingdao

Title: Variants for Visual Odometry Estimation

Visual odometry (VO) has been extensively studied in the last decade.  VO aims at the recovery of a camera trajectory from an image sequence. Stereo vision-based VO techniques solve the egomotion estimation problem by means of disparity-derived 3D scene structure. Typically, one of the two images is only used for disparity computation.
This talk discusses at first the development of a generic feature tracking framework extending the classical VO problem into a higher dimension, where the image data of both cameras are fully used. Six tracking topologies proposed in literature, namely linear, lookahead, stereo linear, parallel, circular and crosseye, are reviewed and evaluated. Based on the experimental results, we found benefits of taking right images into account through the feature tracking process, over the typical stereo VO implementation. The stereo-parallel configuration, which independently maintains feature tracking on each camera and have the tracked features integrated via a left-right matching, has achieved the most significant improvement of 30% over the conventional linear configuration.
Despite a variety of implementation details, proposed VO approaches basically share the same principle - a minimisation of a carefully chosen energy function. This talk also reviews four commonly adopted energy models including perspective, epipolar, rigid, and photometric alignments, and proposes a novel VO technique that unifies multiple objectives for outlier rejection and egomotion estimation to outperform mono-objective egomotion estimation. Experiments show an improvement above 50% is achievable by trading off 15% additional computational costs.  - This talk informs about joint work with Dr. Hsiang-Jen Chien and further (former) students or colleagues.

Reinhard Klette,现任新西兰皇家科学院院士,新西兰奥克兰理工大学教授,机器人与视觉中心主任。1973年和1978年获得德国耶拿大学数学硕士和博士学位,并于1982年获得德国耶拿大学计算机科学博士学位,曾任德国科学院、柏林技术大学,以及新西兰奥克兰大学教授。Reinhard Klette教授在IEEE TCVPR, ICCV, IROS, AI会议发表相关学术论文100余篇。1995年至今,多次应邀作为重要国际学术会议的主讲嘉宾。2003年,于美国马里兰大学Azriel教授合作编著首部综合性专著《数字图像几何算法》。20114月至201310月任JCET(控制工程和技术)期刊创刊主编。2001年至2008年任IEEE PAMI副主编,现任Journal of Information and Communication Convergence Engineering编辑,Computational Imaging and Vision编辑,并任多个期刊编委。欧洲计算机图形和模式分析委员会终身荣誉会员,环太平洋图像和视频技术研讨委员会委员。近期主要研究方向为基于视觉的辅助驾驶、计算机视觉和道路安全、人的姿态理解等。