Combining Edge Images and Depth Maps for Robust Visual Odometry

Abstract

In this work, we propose a robust visual odometry system for RGBD sensors. The
\ncore of our method is a combination of edge images and depth maps for joint camera pose
\nestimation. Edges are more stable under varying lighting conditions than raw intensity
\nvalues and depth maps further add stability in poorly textured environments. This leads
\nto higher accuracy and robustness in scenes, where feature- or photoconsistency-based
\napproaches often fail. We demonstrate the robustness of our method under challenging
\nconditions on various real-world scenarios recorded with our own RGBD sensor.
\nFurther, we evaluate on several sequences from standard benchmark datasets covering a
\nwide variety of scenes and camera motions. The results show that our method performs
\nbest in terms of trajectory accuracy for most of the sequences indicating that the chosen
\ncombination of edge and depth terms in the cost function is suitable for a multitude of
\nscenes.