A Real-Time Robust Visual SLAM Using Dynamic Object Identification and Elimination

Simultaneous Localization and Mapping (SLAM) is a critical technology for estimating the pose of mobile robots and reconstructing maps of the surrounding environment. However, traditional SLAM systems often assume a static environment, resulting in suboptimal performance in dynamic environments. To address this problem, we propose a real-time robust visual SLAM algorithm using YOLOv8, optical flow and blur detection for dynamic target identification and removal. Initially, YOLOv8 extracts semantic information from RGB images and generates bounding boxes for detected objects. Concurrently, ORB feature points are extracted and the optical flow method is used to analyze the motion of these points within the bounding boxes to determine the object's state. However, capturing fast-moving objects results in motion blur, which affects the ability of dynamic SLAM to determine the state of the object. Therefore, we propose a blur detection method to analyze moving blurred targets, thereby enhancing the system's capacity to identify dynamic targets. Finally, feature points from dynamic target boxes are excluded, leaving only remaining static points for pose estimation. Experiments have been conducted on both datasets and real-world scenarios, and the results demonstrate that the proposed algorithm achieves superior performance in dynamic scenes.