Implementation of an RGB Camera-Based Visual SLAM System Using RTAB-Map

Abstract

Simultaneous Localization and Mapping (SLAM) is an essential component in robotic systems and computer vision-based applications for estimating camera motion while constructing an environmental map. This study presents the implementation and qualitative analysis of an RGB camera-based visual SLAM system using RGB images as the sole source of information. The system integrates Visual Odometry to estimate camera motion based on visual feature tracking between consecutive frames, Loop Closure to detect previously visited locations and perform global trajectory correction, and three-dimensional mapping to represent environmental structure. The processing pipeline operates sequentially, starting from image acquisition, visual feature extraction, camera motion estimation, loop closure optimization, and three-dimensional map generation in the form of a point cloud. System evaluation is conducted through descriptive visual analysis of the camera trajectory and the resulting three-dimensional map. The results demonstrate that the proposed visual SLAM system is capable of producing continuous camera motion estimation and visually interpretable three-dimensional maps, although drift accumulation and visual noise are still observed due to the vision-only configuration.