An improved binocular visual odometry for high-speed automotive applications

Yu Huan, Chen Jiabin, Wang Liujun, Xie Ling, Song Chunlei, Wu Qinghe

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

In this paper, we present an improved motion estimation method by adding extra information for binocular visual odometry (VO) which is especially suited for improving high-speed pose change estimation. The extra information is obtained by structured object detecting, taking lane line detection as an example. We can get an accurate position information by calculating the interval of each dotted lane line and counting the number of the dotted line which can be fused with the pose information obtained from visual odometry. The outlier rejection of the VO is also improved, making it adapt to highway situation. In the fusion process, a Kalman filter is adopted to estimate the motion and location information for a high speed vehicle. The experimental results show that the approach proposed is valid and can increase the positioning accuracy significantly compared with ordinary visual odometry.

Original languageEnglish
Title of host publicationProceedings of the 29th Chinese Control and Decision Conference, CCDC 2017
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages185-190
Number of pages6
ISBN (Electronic)9781509046560
DOIs
Publication statusPublished - 12 Jul 2017
Event29th Chinese Control and Decision Conference, CCDC 2017 - Chongqing, China
Duration: 28 May 201730 May 2017

Publication series

NameProceedings of the 29th Chinese Control and Decision Conference, CCDC 2017

Conference

Conference29th Chinese Control and Decision Conference, CCDC 2017
Country/TerritoryChina
CityChongqing
Period28/05/1730/05/17

Keywords

  • Kalman Filter
  • Lane Line Detection
  • Visual Odometry

Fingerprint

Dive into the research topics of 'An improved binocular visual odometry for high-speed automotive applications'. Together they form a unique fingerprint.

Cite this