Abstract:When the existing visual SLAM (simultaneous localization and mapping) algorithms are applied to dynamic environments, the pose error estimated by the system often increases sharply, or even the algorithm fails due to the interference of dynamic objects. To adapt to dynamic scenes, a dynamic object processing part needs to be added to the system. However, some existing processing methods lead to reduced real-time performance, which is not conducive to the real-time localization and navigation of mobile robots. To solve the above problems, an RGB-D SLAM system is proposed in this paper for indoor dynamic environments. The system designs an adaptive semantic segmentation tracking algorithm to meet the requirements of localization accuracy and real-time performance in dynamic scenes. First, a lightweight semantic segmentation network is used to provide a priori information about the object. According to this prior information and the motion state of the object in the previous scene, each feature point is assigned a motion level and is classified as a static point, movable static point, or dynamic point. Then, whether the current frame needs semantic segmentation is adaptively determined according to the motion level information of the feature points. Some appropriate feature points (static points) are selected for initial pose estimation, and then, secondary optimization of the pose is performed according to the results of weighted static constraints. In order to verify the effectiveness of the proposed algorithm, experiments are carried out on the TUM RGB-D dynamic scene dataset and compared with ORB-SLAM2 and other SLAM algorithms for dynamic environments. The results show that the proposed algorithm performs well on most datasets, and the positioning accuracy in indoor dynamic environments can be improved by 90.57% compared with the ORB-SLAM2 algorithm. In addition, a 3D semantic map of static backgrounds in dynamic scenes has been established, using dense point cloud maps to visualize 3D scene information, and incorporating semantic information to label objects in the scene, to guide advanced tasks such as robot navigation and enhance the usability of the system.

Real-time SLAM Algorithm Based on RGB-D Data

Stereo Vision Based SLAM Using Rao-Blackwellised Particle Filter

Robust Keyframe-based Dense SLAM with an RGB-D Camera.

RGB‐D SLAM with Moving Object Tracking in Dynamic Environments

Approach to 3D SLAM for Mobile Robot Based on RGB-D Image with Semantic Feature in Dynamic Environment

Research on RGB-D SLAM Based on Image Feature

An RGBD-SLAM with Bi-directional PnP Method and Fuzzy Frame Detection Module

Real-Time Dense Visual Odometry for RGB-D Cameras

Local Optimized and Scalable Frame-to-model SLAM

Robust and Efficient RGB-D SLAM in Dynamic Environments

A Robust RGB-D SLAM System With Points and Lines for Low Texture Indoor Environments

Simultaneous localization and mapping based on RGB-D images with filter processing and pose optimization

RGB-D SLAM Algorithm Based on Clustering and Geometric Residuals in Dynamic Environment

Development of RGB-D simultaneous localization and mapping in dynamic environments based on motion removal and dense map reconstruction

An RGB-D SLAM algorithm based on adaptive semantic segmentation in dynamic environment

Improved RGB-D vision SLAM algorithm for mobile robot

An efficient SLAM system only using RGBD sensors

Real-Time Optimization-Based Dense Mapping System of RGBD-Inertial Odometry

RVD-SLAM: A Real-Time Visual SLAM Toward Dynamic Environments Based on Sparsely Semantic Segmentation and Outlier Prior

Accurate RGB-D SLAM in Dynamic Environments Based on Dynamic Visual Feature Removal

Towards Real-time Semantic RGB-D SLAM in Dynamic Environments