PmD-ICP: Enhancing LiDAR Localization based on Pseudo-Marker Detection and Reflective Marker-Assisted ICP

Light Detection and Ranging (LiDAR) based autonomous localization, as a core component of intelligent sensing and navigation technologies, has shown significant potential in areas such as autonomous driving, robot navigation, and consumer electronics. As environmental complexity increases, traditional LiDAR localization methods face growing challenges, which has led to the research on combining reflective markers (RMs) with LiDAR localization techniques becoming a research hotspot. However, pseudo-markers, such as high-intensity reflections and marker occlusion, significantly impact LiDAR positioning accuracy. To address these challenges, we propose a novel LiDAR indoor localization method that simultaneously eliminates pseudo-marker interference and marker occlusion problems. Specifically, we introduce a pseudo-marker detection technique based on Riemannian manifolds (PmD-rm) to effectively eliminate interference from the environment. For marker occlusion, a pose estimation optimization algorithm based on Iterative Closest Point (PEO-ICP) is employed to interpolate and optimize the continuous stream of RM point cloud data in occluded environments. The experimental results show that our optimized algorithm improves localization accuracy by approximately 90% in our custom datasets compared to the state-of-the-art (SOTA) algorithms, demonstrating the effectiveness and stability of our method in complex environments.