Detection and utilization of vertical intersection in feature-less environment with RGB-depth sensor

This paper proposes a method to build a RGBD map in feature-less environment. The proposed method especially focuses on a corridor environment a common indoor space composed of simple textureless structures. It estimates robot positions for mapping by using a plane registration algorithm with additional imaginary planes. The imaginary planes are created by vertical intersections the core element of this method. An experimental result shows that the proposed method increases the success rate of estimation of translation along a corridor.