Motion Estimation, 3D Reconstruction and Navigation with Range Sensors

Abstract

In this thesis we leverage range sensors, namely depth cameras and laser scanners, to solve fundamental problems that lie at the interface between computer vision and robotics. Concretely, new algorithms are proposed for sensor tracking (odometry), scene flow estimation, 3D Reconstruction of objects and autonomous navigation of mobile robots. Besides the theoretical contributions described herein, fast implementations of these algorithms are available online for public use.