3D reconstruction system for autonomous robot navigation

Abstract

In this paper, a full 3D reconstruction pipeline is presented to realize the navigation of an autonomous robot. The system utilizes the existing algorithms for multi-view stereo and develops them into an automated system where the input is a video (or random images) obtained by the robot and the output is the model of the reconstructed environment containing the path of the robot exploring it. We have tested the system in several environments and we demonstrate the results that show the potential of this state of the art approach.