Research on Visual SLAM Systems in Indoor Dynamic Scenes.

Saved in:

Bibliographic Details
Title:	Research on Visual SLAM Systems in Indoor Dynamic Scenes.
Authors:	Yanhui Lv¹ yanhuilv@126.com, Qiang Zhong² 1638683463@qq.com
Source:	Engineering Letters. Jan2026, Vol. 34 Issue 1, p164-174. 11p.
Subjects:	Visual odometry, Image segmentation, Statistical reliability
Abstract:	To address the issue of mobile objects in indoor dynamic scenes degrading the localization accuracy of SLAM systems, we propose a visual SLAM system with dynamic point elimination based on ORB-SLAM3. First, an instance segmentation model is designed using YOLOv8-seg, incorporating a GhostBottleneck module to achieve model lightweighting, and introducing the SimAM attention mechanism to enhance segmentation performance. Second, using instance segmentation models to assist SLAM systems in feature extraction, and combining geometric constraint methods to remove dynamic points, in order to improve the pose estimation accuracy of SLAM systems. Finally, a dense point cloud mapping module is developed, where dynamic-area point clouds are filtered using segmentation results, followed by statistical and voxel filtering for optimization. Experimental results demonstrate that compared to ORB-SLAM3, the proposed system reduces the average RMSE of absolute trajectory error by 47.135%. Against other classical SLAM systems, it exhibits superior dynamic point removal capability in dynamic environments and effectively mitigates the impact of dynamic objects on dense point cloud map reconstruction. [ABSTRACT FROM AUTHOR]
	Copyright of Engineering Letters is the property of International Association of Engineers (IAENG) and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)
Database:	Engineering Source

Description
Abstract:	To address the issue of mobile objects in indoor dynamic scenes degrading the localization accuracy of SLAM systems, we propose a visual SLAM system with dynamic point elimination based on ORB-SLAM3. First, an instance segmentation model is designed using YOLOv8-seg, incorporating a GhostBottleneck module to achieve model lightweighting, and introducing the SimAM attention mechanism to enhance segmentation performance. Second, using instance segmentation models to assist SLAM systems in feature extraction, and combining geometric constraint methods to remove dynamic points, in order to improve the pose estimation accuracy of SLAM systems. Finally, a dense point cloud mapping module is developed, where dynamic-area point clouds are filtered using segmentation results, followed by statistical and voxel filtering for optimization. Experimental results demonstrate that compared to ORB-SLAM3, the proposed system reduces the average RMSE of absolute trajectory error by 47.135%. Against other classical SLAM systems, it exhibits superior dynamic point removal capability in dynamic environments and effectively mitigates the impact of dynamic objects on dense point cloud map reconstruction. [ABSTRACT FROM AUTHOR]
ISSN:	1816093X