基于动态特征筛选的室内服务机器人视觉SLAM改进

Enhanced Visual SLAM for Indoor Service Robots via Dynamic Feature Filtering

摘要: 针对室内服务机器人在动态环境下视觉SLAM (Simultaneous Localization and Mapping)系统鲁棒性不足的问题，本文提出了一种基于动态特征筛选的改进方法。通过融合RGB-D相机与惯性测量单元(IMU)数据，结合语义分割与运动一致性检测，动态剔除环境中的移动物体特征点，优化特征匹配与位姿估计流程。实验结果表明，在包含行人、移动障碍物等动态干扰的室内场景中，改进后的SLAM系统定位误差较传统ORB-SLAM3降低42%，地图重建完整性提升28%。

Abstract: To address the issue of insufficient robustness of the visual SLAM (Simultaneous Localization and Mapping) system for indoor service robots in dynamic environments, this paper proposes an improved method based on dynamic feature screening. By integrating RGB-D camera and inertial measurement unit (IMU) data, and combining semantic segmentation and motion consistency detection, the method dynamically eliminates the feature points of moving objects in the environment, optimizing the feature matching and pose estimation processes. Experimental results show that in indoor scenes with dynamic disturbances such as pedestrians and moving obstacles, the positioning error of the improved SLAM system is reduced by 42% compared to the traditional ORB-SLAM3, and the map reconstruction completeness is increased by 28%.