Some improvements in the RGB-D SLAM system

Abstract

RGB-D cameras offer both color and depth images of the surrounding environment, making them an attractive option for robot sensor. In this work, we present an RGB-D SLAM system using the Microsoft Kinect. The proposed system is a full 6DoF (Degrees of Freedom) SLAM system which can estimate camera trajectory and reconstruct a 3D model of the environment in real-time. Unlike traditional filtering-based approaches, our system optimizes the entire trajectory by the use of graph optimization. We achieve better accuracy than a previous system by employing key-frame matching instead of frame-to-frame matching. We evaluate our system on a published dataset. The results demonstrate that our system can handle unrestricted camera movements in indoor settings.