Kinect-based 3D Indoor Environment Map Building

Abstract

For the problem of poor robustness and real-time performance of mobile robot during the process of 3D map building, a new 3D map building method which combines the improved EKF-based RANSAC algorithm and the G-ICP algorithm is proposed. First of all, a feature point extraction and matching process is carried out based on the RGB data collected by Kinect. During the process of matching, the improved RANSAC algorithm is employed to eliminate the false matching points. Then, in order to get the three-dimensional point cloud data of the indoor environment, a G-ICP algorithm which has better robustness and performance is employed to accomplish accurate image registration. Eventually, a 3D indoor environment map is created by the matching data and 3D point cloud data. The effectiveness of the proposed method is verified by the experiments.