Evaluation of a Stereo Visual Odometry Algorithm for Road Vehicle Navigation

Paper #:
  • 2017-01-0046

Published:
  • 2017-03-28
Abstract:
To reliably implement driver-assist features and ultimately, self-driving cars, autonomous driving systems will likely rely on a variety of sensor types including GPS, RADAR, LASER range finders, and cameras. Cameras are an essential sensory component because they lend themselves to the task of identifying object types that a self-driving vehicle is likely to encounter such as pedestrians, cyclists, animals, other cars, or objects on the road. A stereo vision system adds the capability of tracking object locations and trajectories relative to the vehicle. This information can be essential for an autonomous driving control system that aims to avoid collisions and localize itself in the street scene. In this paper, we present a visual odometry algorithm based on a stereo-camera to perform localization relative to the surrounding environment for purposes navigation and hazard avoidance. Using a stereo-camera enhances the accuracy with respect to monocular visual odometry. It avoids the bootstrapping problem and enables the estimation of the exact trajectory in metric space, which is important for proper position control and path planning. The implemented algorithm is feature-based which relies on detecting AGAST corners and tracking them across frames. BRIEF descriptors are computed and used for matching the detected corners. These features are used in stereo-correspondence in order to triangulate and track local 3D map points. The established 3D-2D matches are used in motion estimation by minimizing the re-projection error. This localization approach alongside readily-available stereo disparity maps can be used for obstacle detection and navigation. The visual odometry algorithm is evaluated using the widely used KITTI dataset and benchmarking suite (project by Karlsruhe Institute of Technology and Toyota Technological Institute). The algorithm’s accuracy is evaluated, and specific scenarios in the dataset that challenge the algorithm are discussed to identify potential areas for future research towards vehicle full-autonomy.
Access
Now
SAE MOBILUS Subscriber? You may already have access.
Buy
Attention: This item is not yet published. Pre-Order to be notified, via email, when it becomes available.
Select
Price
List
Download
$22.00
Mail
$22.00
Members save up to 36% off list price.
Share
HTML for Linking to Page
Page URL

Related Items

Technical Paper / Journal Article
2003-06-17
Book
2009-03-01
Article
2016-02-02
Standard
2006-09-12
Article
2016-02-02
Article
2016-02-02
Training / Education
2017-04-03