Improved KinectFusion based on graph-based optimization and large loop model

Abstract

In dense 3D SLAM (simultaneous localization and mapping), the use of RGB-D data to realize SLAM has become more widespread. In this paper, PRKF (the precise and robust KinectFusion) is proposed. On the basis of KinectFusion, the graph optimization based on g2o (general graph optimization) is added to the PRKF. In the g2o optimization policy, in order to achieve the rapid optimization of error accumulation, this paper proposes a model based on the registration model for model loop optimization. The Kinect sensor is carried by the Pioeer3-DX to establish the map of LAB in real time. In addition, public data sets FR1 is used to compare KinectFusion with the PRKF in this paper. The experiments have proved that the algorithm is robust and has high precision.