Visual odometry with ground facing camera

Speaker: Merwan Birem

Company/ Institute: Flanders Make

Event: 20th work meeting

Date: 8 December 2015

Abstract:

Odometry is the use of data from motion sensors to estimate change in position over time. Odometry is often used on mobile robots and vehicles to estimate their position relative to a starting location. As an alternative to classical odometry based on Inertial Measurement Units (IMUs), a novel visual odometry method based on a ground facing camera is proposed. The approach is based on the Fourier transform, which is used to determine the precise translation distances (for 2D position) and rotation angles (for heading) between consecutive images captured by a ground facing camera.  The proposed method is especially suited if no distinct visual features are present on the ground. The method has been tested on different experimental platforms and evaluated against the ground truth,  including a successful loop-closing test, to demonstrate its general use and performance. Flanders Make will show the results obtained for the loop-closure test, and give a live demo using a trolley.

Slides: available after the workshop (Community members only, login required)