基于YOLOv5的移动机器人动态视觉SLAM算法研究

Research on Dynamic Visual SLAM Algorithm for Mobile Robots Based on YOLOv5

移动机器人在未知环境中,通过同步定位与地图构建(SLAM)技术,实现了精准的自身定位功能。目前大多数视觉SLAM系统均假设环境是静态的,但在实际应用中,由于大量动态目标的存在,严重影响机器人的定位与建图精度。为改善这一情况,本文基于ORB-SLAM3系统提出一种鲁棒的动态视觉SLAM系统,其融合YOLOv5深度学习方法,以减少动态目标的影响。并在公共TUM数据集和真实场景中测试本文算法的性能,结果表明:本文算法与ORB-SLAM3相比,具有更高的鲁棒性。

Mobile robots can achieve precise self-localization through Simultaneous Localization and Mapping(SLAM)technology in unknown environments.Most current visual SLAM systems assume that the environment is static,but in practical applications,the presence of a large number of dynamic objects seriously affects the robot’s localization and mapping accuracy.To improve this situation,this paper proposes a robust dynamic visual SLAM system based on the ORB-SLAM3 system,which integrates the YOLOv5 deep learning method to reduce the impact of dynamic objects.The per-formance of the algorithm in this paper is tested on the public TUM dataset and real-world scenarios,and the results show that the algorithm in this paper has higher robustness compared with ORB-SLAM3.