Point-Plane SLAM Using Supposed Planes for Indoor Environments

Abstract

Simultaneous localization and mapping (SLAM) is a fundamental problem for various applications. For indoor environments, planes are predominant features that are less affected by measurement noise. In this paper, we propose a novel point-plane SLAM system using RGB-D cameras. First, we extract feature points from RGB images and planes from depth images. Then plane correspondences in the global map can be found using their contours. Considering the limited size of real planes, we exploit constraints of plane edges. In general, a plane edge is an intersecting line of two perpendicular planes. Therefore, instead of line-based constraints, we calculate and generate supposed perpendicular planes from edge lines, resulting in more plane observations and constraints to reduce estimation errors. To exploit the orthogonal structure in indoor environments, we also add structural (parallel or perpendicular) constraints of planes. Finally, we construct a factor graph using all of these features. The cost functions are minimized to estimate camera poses and global map. We test our proposed system on public RGB-D benchmarks, demonstrating its robust and accurate pose estimation results, compared with other state-of-the-art SLAM systems.

Feature-based visual simultaneous localization and mapping: a survey

“…A recent RGB-D SLAM was proposed in [144] where points and planes are exploited to estimate the pose of a camera and a map of its surroundings. ORB features are extracted from RGB frames and handled by the RGB-D version of ORB-SLAM2.…”

RGB-D Object SLAM Using Quadrics for Indoor Environments

“…Visual SLAM systems observe landmarks from different poses and construct constraints to solve the camera poses and the landmarks’ locations [ 1 ]. Conventional visual SLAM systems generally use point features [ 8 , 9 , 20 ], line features [ 12 ], and plane features [ 10 , 11 ] as landmarks. Those low-dimensional geometric representations can help the robot locate its poses, while the lack of semantic information in the map limits the mobile robot’s ability to understand the environment.…”

“…Conventional visual SLAM uses the descriptor of point features [ 6 , 7 ], or the geometric difference of line and plane features [ 10 , 11 , 12 ], to track and solve data associations of observations between frames. After associations solved at the front end, it is fixed during the optimization in the back end.…”

“…A life-long robot needs to adapt to illumination changes, sensor noise, and significant viewing angle changes—all of those put forward requirements for the robustness of landmarks in its map. The existing visual SLAM technology relies on feature descriptors [ 6 , 7 ], surface textures [ 8 , 9 ], or the geometric position of plane and line structures [ 10 , 11 , 12 ] to solve data associations. However, artificially designed descriptors are challenging to adapt to significant viewing angle changes and are easily disturbed by illumination and sensor noise.…”

Depth-Image Segmentation Based on Evolving Principles for 3D Sensing of Structured Indoor Environments

“…Modern technologies enable the development of new and more sophisticated sensors, such as stereo cameras, RGB-D sensors, Laser Range Finders (LRF), etc., which are often equipped with powerful processor units to capture and process information-rich data in real time. Due to their increased popularity and ubiquity, they are indispensable in mobile robotics applications [ 1 , 2 , 3 , 4 ]. Processing the data obtained from the aforementioned sensors is a computationally demanding task; therefore, it is necessary to ensure the optimal method of processing the acquired data for real-time operations is used, such as vision-based high speed driving [ 5 ] or visual inspection in manufacturing processes [ 6 ].…”