Fiducial Tag Localization on a 3D LiDAR Prior Map

TL;DR

The paper addresses localizing LiDAR fiducial tags directly on a 3D LiDAR prior map to enable light-insensitive, map-aware navigation. It introduces a joint intensity-geometry pipeline that downscales the point cloud via 3D intensity gradients, clusters prospective tag regions with oriented bounding boxes, filters candidates using size and shape criteria, and uses an intermediate-plane projection to perform IFM-based tag detection and 6-DOF pose estimation. Key contributions include the first method for 3D-map tag localization, a two-stage filtering strategy to isolate tag-like clusters, and an intermediate-plane approach that improves imaging quality and pose accuracy, all released in open-source form. This work enables robust LiDAR-based relocalization and navigation in environments where visual fiducials are unreliable, by directly leveraging 3D map information and LiDAR-specific tag characteristics.

Abstract

The LiDAR fiducial tag, akin to the well-known AprilTag used in camera applications, serves as a convenient resource to impart artificial features to the LiDAR sensor, facilitating robotics applications. Unfortunately, the existing LiDAR fiducial tag localization methods do not apply to 3D LiDAR maps while resolving this problem is beneficial to LiDAR-based relocalization and navigation. In this paper, we develop a novel approach to directly localize fiducial tags on a 3D LiDAR prior map, returning the tag poses (labeled by ID number) and vertex locations (labeled by index) w.r.t. the global coordinate system of the map. In particular, considering that fiducial tags are thin sheet objects indistinguishable from the attached planes, we design a new pipeline that gradually analyzes the 3D point cloud of the map from the intensity and geometry perspectives, extracting potential tag-containing point clusters. Then, we introduce an intermediate-plane-based method to further check if each potential cluster has a tag and compute the vertex locations and tag pose if found. We conduct both qualitative and quantitative experiments to demonstrate that our approach is the first method applicable to localize tags on a 3D LiDAR map while achieving better accuracy compared to previous methods. The open-source implementation of this work is available at: https://github.com/York-SDCNLab/Marker-Detection-General.

Figures (15)

• Figure 1: An illustration of the proposed method, which jointly analyzes the point cloud of the 3D map from both intensity and geometry perspectives, returning the tag poses and vertex locations w.r.t. the global coordinate system of the map. (a): After sections \ref{['ic']} and \ref{['clu']}. (b): After section \ref{['clu2']}. As seen, all the clusters but the one belonging to the ArUco tag aruco are filtered out. (c): The zoomed view of the preserved oriented bounding box. (d): A zoomed view of the detected 3D fiducials, which are rendered in red.
• Figure 2: The 3D map shown in Fig. \ref{['mov']} observed from the origin of the global coordinate system. The occlusion is not an artificial challenge, and a detailed explanation is provided in Remark 1.
• Figure 3: The example used to explain the design purpose and result of each step. The two presenters are holding two different AprilTags ap3. As illustrated in the top view, observing along the X-axis of the global coordinate system, the back subpoint cloud is totally blocked by the front one. Thus, Eq. (\ref{['pro']}) cannot function in this case.
• Figure 4: The effect of applying downsampling from the intensity perspective. As seen, the majority of unnecessary points are filtered out at this step.
• Figure 5: A diagram to illustrate the design of a typical square fiducial tag ap3. A square fiducial tag is a combination of the prototype tag (a black frame inside a white frame) and the encoding area.