RGB-D SLAM in Indoor Environments With STING-Based Plane Feature Extraction

RGB-D SLAM in Indoor Environments With STING-Based Plane Feature Extraction

Abstract

In this paper, the RGB-D camera-based simultaneous localization and mapping (SLAM) of indoor environments is addressed using plane features. The plane parameter space (PPS) is defined for a compact representation of planes in the Cartesian space. The statistical information grid (STING) structure is constructed in the PPS to extract plane features. The plane association graph is developed to determine the correspondences between the plane features from two successive scans. The RGB-D camera pose is directly calculated using the matched plane features if they can provide sufficient constraints for the pose estimation. Otherwise, a novel STING-based scan matching method is developed in the PPS to achieve a full six degrees of freedom camera pose estimation. The proposed method uses only the plane features independent of any other features to estimate the RGB-D camera poses and can thus be suitable for some challenging scenes. The experimental results demonstrate that the proposed plane feature-based RGB-D SLAM can achieve high accuracy and robustness in both on-board and hand-held applications.

Keywords: Indoor environment mapping, plane feature, RGB-D camera, robot vision, six degrees of freedom (6-DoF) camera pose estimation

The video shows the online robot SLAM experiment in a real world scene presented in Section V.D of the paper. Specifically, it shows the process of the RGB-D visual odometry and incremental mapping achieved by the proposed plane feature-based RGB-D camera pose estimation (PF-RGBD-CPE) method.

More in details:

• Experimental Setup: The size of the laboratory is approximately 12.0m×5.2m. The Kinect is mounted on the Pioneer 3-DX mobile robot 1.14m above the ground, pointing to the right side of the robot. The PF-RGBD-CPE runs on an onboard computer (Intel Pentium Dual T2390 CPU at 1.86GHz, 3G RAM). The length of the trajectory is approximately 46.7m, covering two loops around the laboratory. During the first loop, the Kinect points to the tables in the middle of the room, and during the second loop it points to the surrounding walls.

• Video Contents:

  • Text on the top: frame number, number of extracted planes and the constraint case of the current alignment;
  • RGB Image & Depth Image: RGB and depth images collected by the Kinect sensor;
  • Camera 1 & Camera 2: views of the two fisheye cameras fixed on the ceiling (two cameras are used to cover the whole room);
  • Point Cloud Map: the 3D point cloud map as well as the trajectory of the RGB-D sensor estimated by PF-RGBD-CPE;
  • Extracted Planes: the extracted planes in the map shown in different colors.

Recommended citation: Q. Sun, J. Yuan, X. Zhang, F. Sun. RGB-D SLAM in Indoor Environments With STING-Based Plane Feature Extraction. IEEE/ASME Transactions on Mechatronics, 2018, 23(3): 1071-1082. http://sunqinxuan.github.io/files/publications-2018-06-12-TMech.pdf