Pith. sign in

REVIEW 3 major objections 4 minor 96 references

Mapping and Localization Using LiDAR Fiducial Markers

T0 review · 3 major / 4 minor · reviewed 2026-08-09 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read Thin-sheet LiDAR fiducial markers let unordered, low-overlap point clouds be registered into accurate maps.

desk verdict Solid integration of prior LiDAR-fiducial work with a useful dataset, but the 'any unseen scenario' claim needs scoping: the method depends on co-observed markers forming a connected graph. read the letter →

arxiv 2502.03510 v1 pith:SSZHWCU2 submitted 2025-02-05 cs.CV

Yibo Liu This is my paper · ORCID
classification cs.CV
keywords LiDARfiducialmarkersintensityimagepointcloudregistrationmultiviewlowoverlapscansmappingandlocalizationfactorgraphtrainingdatacollection
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This dissertation sets out to close the gap between visual fiducial markers, which are common in camera-based systems, and LiDAR fiducial markers, which are not. It claims that thin, letter-sized printed markers detected through LiDAR intensity images can serve as reliable anchors for registering unordered multiview point clouds, even when the scans share almost no geometric features. If the claim holds, tasks such as collecting 3D assets from a few sparse scans, merging large-scale maps with only a few percent overlap, reconstructing degraded scenes, and localizing without GPS become feasible with paper markers and a LiDAR. The central mechanism is a two-level graph that first initializes scan poses from shared marker observations and then jointly optimizes scan poses, marker poses, and marker corners. The strongest stated consequence is that the framework works in unseen scenarios with extremely low overlap, exactly where geometry-based and learning-based registration methods fail.

What carries the argument

The central object is the Intensity Image-based LiDAR Fiducial Marker (IFM) system: letter-sized printed markers carrying ordinary visual fiducial patterns are projected from raw LiDAR intensity into a spherical intensity image, a standard visual-fiducial detector locates the 2D fiducials, and an interpolation scheme converts detected but unscanned feature points into 3D coordinates using the planar-marker assumption. On top of this sit two further mechanisms. An adaptive threshold detector searches the binarization threshold to maximize the number of markers detected as viewpoints change. A two-level graph then carries the registration: the first-level graph is a weighted graph whose nodes are scans and markers and whose edge weights are the point-to-point pose errors $e_{pp}$; Dijkstra's algorithm finds the lowest-error path from the anchor scan to each non-anchor scan to initialize all poses. The second-level graph is a factor graph that formulates the maximum a-posteriori problem, the most probable poses given the measurements, and jointly optimizes scan poses, marker poses, and marker corner positions via Levenberg-Marquardt.

What would settle it

Take two LiDAR scans of the same room with no marker visible in the second scan, or with the second scan seeing only markers absent from the first, and run the full pipeline to register them; the first-level graph will have no path from the anchor scan to that scan, so the claimed registration of any unseen low-overlap input cannot be produced unless the method silently falls back on geometry.

Watch

Extended reading notes

Core claim

The discovery is that LiDAR fiducial markers do not have to be bulky three-dimensional objects. By projecting a LiDAR point cloud onto a spherical intensity image, a standard visual fiducial detector can find the marker corners, and those corners give enough constraints to estimate the sensor pose. The same idea is extended from single scans to 3D maps by analyzing the point cloud jointly from intensity and geometry, using an intermediate plane to remove occlusion and enlarge small distant markers. Multiview registration is then posed as a maximum a-posteriori problem over scan poses, marker poses, and corner positions, solved by a first-level weighted graph that propagates poses along lowest-error paths from an anchor scan and a second-level factor graph that refines all variables globally. In the author's telling, this is why the method registers low-overlap and scene-degraded inputs that defeat feature-based competitors, and why it can serve as an efficient tool for map merging, asset collection, and training-data generation.

Load-bearing premise

The load-bearing premise is that every scan to be registered shares at least one marker with some other scan, so the first-level graph can find a path from the anchor scan to that scan; the dissertation does not test scans that see no markers or only markers invisible to all other scans.

Editorial extensions

If this is right

  • Mapping and localization no longer depend on rich geometric features in the overlap region, so scans whose shared area is mostly planar or textureless can still be aligned.
  • Large-scale 3D maps with only a few percent overlap can be merged using marker detections localized on the maps themselves.
  • A LiDAR-only setup can support offline, structure-from-motion-style processing of unordered scans, enabling 3D asset collection from a handful of sparse viewpoints.
  • Training data for learning-based registration can be collected in unseen indoor and outdoor scenes, and adding that data improves the registration recall of existing methods on standard benchmarks.
  • Because the pipeline registers through markers rather than whole-point-cloud feature analysis, its runtime is orders of magnitude shorter than the compared full-analysis methods.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The real scaling limit of the approach is marker co-visibility: a scan that sees no shared markers cannot be initialized by the first-level graph, so the practical operating range is bounded by marker density and LiDAR range rather than by scene geometry.
  • The same factor-graph formulation could accept camera-based visual fiducial detections as additional measurements, which would let the system fall back on visual markers in weather or lighting conditions where LiDAR intensity is degraded.
  • A natural testable extension is to replace the adaptive threshold search with a learned detector that predicts marker corners directly from the raw intensity image, which would remove the threshold loop and make the pipeline faster and more continuous.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 4 minor

Summary. The dissertation develops an intensity-image-based LiDAR fiducial marker (IFM) system, extends marker localization from single scans to 3D maps, and proposes a two-level graph framework for registering unordered low-overlap multiview point clouds using thin-sheet markers. It also introduces the Livox-3DMatch dataset for training learning-based registration methods and reports experiments against MDGD, SGHR, SE3ET, GeoTrans, Teaser++, LOAM variants, KISS-ICP, and SfM-M across indoor/outdoor scenes, sparse-scan reconstruction, degraded scenes, GPS-denied localization, and large-scale map merging. The central claim is that this marker-based framework is robust to any unseen scenarios with extremely low overlap, providing a convenient, efficient, and low-cost mapping/localization tool.

Significance. The integrated system is a useful practical contribution: it transfers the convenience of visual fiducial markers to LiDAR point clouds, provides open-source implementations and datasets, and demonstrates large accuracy and runtime advantages over learning-based registration baselines in scenes instrumented with markers. The SVD-based pose estimation, the factor-graph MAP formulation, and the dataset-augmentation idea are each standard, but their combination is applied to a relevant and previously underexplored problem. The quantitative gains in the reported applications, such as the improvements to SGHR and MDGD from Livox-3DMatch (Table 4.4) and the low map-merging error in Table 4.7, are concrete and useful if the results are reproducible. The main weakness is that the headline robustness claim is broader than the algorithmic conditions and experimental protocol support.

major comments (3)
  1. [Section 4.1 and Section 4.2.3, Eq. (4.3)] The claim that the framework is "robust to any unseen scenarios with extremely low overlap" (Section 4.1, repeated in Section 4.4.3) is not supported by the algorithm as stated. In the first-level graph, a non-anchor scan receives an initial pose only if a path of co-observed markers connects it to the anchor scan; Eq. (4.3) requires a marker observed in both scans, and Dijkstra's algorithm operates over that co-visibility graph. A scan that sees no marker, or only markers not detected in any other scan, remains an isolated node and is never initialized, so the second-level graph cannot register it. None of the ten scenes in Table 4.2 or the applications in Sections 4.4.4-4.4.8 contains a disconnected or marker-free scan. The paper should either add experiments with disconnected marker co-visibility (for example, a scan with no detected markers, or a scan whose markers are observed by no other scan) or explicitly restrict the claim to scenes where the marker co-visibility graph is connected and every scan observes at least one marker.
  2. [Section 4.3 and Section 4.4.5] The Livox-3DMatch dataset is created using poses generated by the proposed marker-based pipeline itself, and no external ground truth (MoCap, RTK, or manually verified CAD alignment) is reported for the aligned 33 scans. The observed improvements to SGHR and MDGD on 3DMatch, ETH, and ScanNet are evidence that the added data are useful for training, but they do not establish the absolute accuracy of the generated poses. Since the dataset is presented as a training resource, the paper should either report a subset of held-out scans with independent pose ground truth or explicitly state that the poses inherit the accuracy of the proposed pipeline and cannot be treated as independently validated ground truth.
  3. [Sections 4.2.1, 3.2.3, and 4.2.4] Several free parameters that affect the reported results are not specified in the manuscript: the search scope S and step size δ in Algorithm 1, the OBB amplification factor t_b and the filtering thresholds in Section 3.2.3, and the noise covariance matrices Σ_k in Eq. (4.6), which are said to be determined by experiments in [39] without reporting the actual values. Without these settings, the quantitative results in Tables 4.2, 4.5, 4.6, and 4.7 cannot be reproduced from the text. Please provide the parameter values, and ideally a sensitivity analysis for the most important ones, or a clear statement that the open-source code contains the exact settings.
minor comments (4)
  1. [Throughout] There are several typographical issues: "T race" in Eq. (2.6) and nearby text should be "Trace"; "Apirltag" in Section 1.2.1 should be "AprilTag"; and "synthesis point cloud" in Section 3.2 should be "synthetic point cloud."
  2. [Section 4.2.1, Algorithm 1] In Algorithm 1, λ* is updated whenever Qtemp,l >= Ql, even if no new marker was appended to Q. This means the returned threshold is the last threshold with at least the current maximum number of detections, not necessarily the threshold at which the final marker set was first detected. Please clarify whether this is intentional and update the pseudocode or explanation.
  3. [Section 4.4.1] The ground-truth poses for the registration-accuracy experiments are obtained by manual alignment in CloudCompare. Given that the reported translation errors are often only a few centimeters, it would be helpful to quantify the precision of the manual ground truth, for example by cross-checking a subset of scenes against MoCap or RTK poses, or by reporting inter-annotator variability.
  4. [Table 3.1] The scene labels in Table 3.1 refer to figure panels ("Fig. 3.1(a)-Traj LO") rather than descriptive scene names. Using explicit scene names would make the table easier to read and compare.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the central mapping/localization claims are anchored to external ground truth, independent benchmarks, and code-released prior work.

full rationale

The derivation chain is self-contained in the relevant sense. Marker detection starts from raw LiDAR intensity values and known VFM pattern geometry, and pose estimation is a standard SVD least-squares alignment of detected 3D fiducials to predefined vertices; this is validated against OptiTrack MoCap ground truth (Section 2.5.1) and against LiDARTag on an external rosbag (Table 2.2). The multiview registration pipeline in Chapter 4 obtains initial relative poses from Eq. (4.3), which uses marker-pose measurements rather than any fitted target pose, and the second-level factor graph optimizes those same measurements with covariances taken from the author's previously published, open-source IFM work [39]. That self-citation is not load-bearing circularity because [39] is externally published, code-released, and does not encode the registration results being claimed. The Livox-3DMatch dataset is aligned using the proposed pipeline, but its value is demonstrated by improved benchmark performance of independent learning methods (Table 4.4), while the proposed method's own pose accuracy is separately checked against manually registered CloudCompare ground truth, MoCap, and RTK (Tables 4.2, 4.5, 4.6), so the reported success does not reduce to the method's own outputs. The 'any unseen scenarios' claim in Section 4.1 is broader than the tested regime because the first-level graph requires a path of co-observed markers to the anchor scan (Eq. 4.3), but that is an overgeneralization or scope limitation, not a circular derivation: no equation in the paper defines the claimed robustness in terms of the experimental outcome, and no fitted parameter is renamed as a prediction. I therefore find no step that can be quoted as reducing a claimed result to its own input by construction.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

The central derivation uses standard SVD pose estimation and factor-graph optimization; the main dependencies are empirical marker detectability, marker co-observability, manual ground-truth precision, and a set of hand-chosen filtering parameters.

free parameters (4)
  • Adaptive threshold search parameters S and step size δ = unspecified
    Algorithm 1 in Section 4.2.1 requires S and δ, but the paper gives no values and no sensitivity analysis; the λ* values in Table 4.1 vary from 6 to 28, showing detection depends strongly on these parameters.
  • OBB buffer amplification factor t_b = twice the marker side length
    Section 3.3 says the recommended value is twice the marker side length based on the authors' experiments; no principled derivation or ablation of t_b is provided.
  • OBB filtering thresholds = SOBB in [a^2, 2a^2]; 1/1.5 <= l/w <= 1.5
    Section 3.2.3 uses manually chosen geometric bounds to accept or reject candidate clusters; the 1.5 ratio and diagonal bounds are not derived from a noise model.
  • Noise covariance matrices Σ_k = from experiments in [39]
    Section 4.2.4 states the covariances are determined by quantitative experiments in the author's prior paper [39], but the values and calibration procedure are not reproduced.
assumptions (5)
  • domain assumption Markers are planar thin sheets whose black-and-white patterns produce locally linear intensity gradients in LiDAR intensity images.
    The whole marker detection pipeline in Sections 2.3 and 3.2 assumes intensity contrast and planarity; the paper itself notes failure at 50 m due to sparse and noisy points.
  • domain assumption The spherical projection with Θa and Θi chosen from the LiDAR manual preserves enough pattern structure for VFM decoders after thresholding and blurring.
    Section 2.3.2 establishes image generation but does not prove marker detectability; CCTag required Gaussian blur to work, showing fragility.
  • domain assumption Every non-anchor scan is connected to the anchor scan through a path of co-observed markers in the first-level graph.
    Section 4.2.3 uses Dijkstra on a graph of shared marker observations; scans with no shared-marker path have no initial pose.
  • standard math Standard SVD least-squares pose result uniqueness for coplanar non-collinear points applies and eliminates rotation ambiguity.
    Section 2.4 relies on this lemma from Arun et al. 1987 for closed-form pose and uniqueness.
  • domain assumption Ground truth poses obtained by manual alignment in CloudCompare are accurate enough for the error levels reported.
    Used in Sections 3.5.1 and 4.4; manual annotation has unknown precision and is not independently verified.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Mapping and Localization Using LiDAR Fiducial Markers." pith.science (2026). https://pith.science/paper/SSZHWCU2

@misc{pith2026250203510,
  author       = {Pith},
  title        = {Pith review of: Mapping and Localization Using LiDAR Fiducial Markers},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/SSZHWCU2}},
  note         = {Machine review of arXiv:2502.03510}
}
read the original abstract

LiDAR sensors are essential for autonomous systems, yet LiDAR fiducial markers (LFMs) lag behind visual fiducial markers (VFMs) in adoption and utility. Bridging this gap is vital for robotics and computer vision but challenging due to the sparse, unstructured nature of 3D LiDAR data and 2D-focused fiducial marker designs. This dissertation proposes a novel framework for mapping and localization using LFMs is proposed to benefit a variety of real-world applications, including the collection of 3D assets and training data for point cloud registration, 3D map merging, Augmented Reality (AR), and many more. First, an Intensity Image-based LiDAR Fiducial Marker (IFM) system is introduced, using thin, letter-sized markers compatible with VFMs. A detection method locates 3D fiducials from intensity images, enabling LiDAR pose estimation. Second, an enhanced algorithm extends detection to 3D maps, increasing marker range and facilitating tasks like 3D map merging. This method leverages both intensity and geometry, overcoming limitations of geometry-only detection approaches. Third, a new LFM-based mapping and localization method registers unordered, low-overlap point clouds. It employs adaptive threshold detection and a two-level graph framework to solve a maximum a-posteriori (MAP) problem, optimizing point cloud and marker poses. Additionally, the Livox-3DMatch dataset is introduced, improving learning-based multiview point cloud registration methods. Extensive experiments with various LiDAR models in diverse indoor and outdoor scenes demonstrate the effectiveness and superiority of the proposed framework.

Figures

Figures reproduced from arXiv: 2502.03510 by the authors.

Figure 1.1
Figure 1.1. Comparison of (a) a typical calibration board, (b)(c) VFM, (d) LiDARTag, [PITH_FULL_IMAGE:figures/full_fig_p035_1_1.png] view at source ↗
Figure 2.1
Figure 2.1. An illustration of the flexibility and generalizability of the proposed IFM. [PITH_FULL_IMAGE:figures/full_fig_p043_2_1.png] view at source ↗
Figure 2.2
Figure 2.2. Comparison of LiDARTag [9] (top) and the proposed IFM (bottom). 2.2 Preliminaries 2.2.1 Three-dimensional Transformation Suppose that pa = [x, y, z] T is a 3D point expressed in the coordinate system {a}. To express the point in another coordinate system {b} as pb = [x ′ , y′ , z′ ] T , translation and 18 [PITH_FULL_IMAGE:figures/full_fig_p045_2_2.png] view at source ↗
Figures from the paper (42 more)
Figure 2.3
Figure 2.3. Figure 2.3: The schematic diagram of the working principle of LiDAR. [PITH_FULL_IMAGE:figures/full_fig_p048_2_3.png]
Figure 2.4
Figure 2.4. Figure 2.4: An illustration of the coordinate systems and notations. [PITH_FULL_IMAGE:figures/full_fig_p049_2_4.png]
Figure 2.5
Figure 2.5. Figure 2.5: The intensity images generated under different angular resolution settings. [PITH_FULL_IMAGE:figures/full_fig_p052_2_5.png]
Figure 2.6
Figure 2.6. Figure 2.6: Sampling patterns of the mechanical LiDAR and solid-state LiDAR, with [PITH_FULL_IMAGE:figures/full_fig_p053_2_6.png]
Figure 2.7
Figure 2.7. Figure 2.7: An illustration of image preprocessing in the system. [PITH_FULL_IMAGE:figures/full_fig_p054_2_7.png]
Figure 2.8
Figure 2.8. Figure 2.8: An illustration of the algorithm to estimate the 3D coordinates of a detected [PITH_FULL_IMAGE:figures/full_fig_p056_2_8.png]
Figure 2.9
Figure 2.9. Figure 2.9: The side view of α. ϕd, ϕk, and ϕu are the inclinations of pd, pk, and pu, respectively. Undoubtedly, OLpk is the angle bisector of ∠puOLpd owing to ϕd − ϕk = ϕk − ϕu. Hence, in the light of the angle bisector properties, we have pkpd/pupk = OLpd/OLpu. Note that OLpd…
Figure 2.10
Figure 2.10. Figure 2.10: An illustration of the experimental setup. [PITH_FULL_IMAGE:figures/full_fig_p063_2_10.png]
Figure 2.11
Figure 2.11. Figure 2.11: Marker detection results on the preprocessed intensity images. [PITH_FULL_IMAGE:figures/full_fig_p064_2_11.png]
Figure 2.12
Figure 2.12. Figure 2.12: Pose estimation accuracy of the IFM system and the AprilTag 3 system at [PITH_FULL_IMAGE:figures/full_fig_p066_2_12.png]
Figure 2.13
Figure 2.13. Figure 2.13: An illustration of the limitation of spherical projection for 3D maps. This [PITH_FULL_IMAGE:figures/full_fig_p071_2_13.png]
Figure 3.1
Figure 3.1. Figure 3.1: An overview of the improvements to the vanilla IFM. [PITH_FULL_IMAGE:figures/full_fig_p074_3_1.png]
Figure 3.2
Figure 3.2. Figure 3.2: The example used to explain the design purpose and result of each step. [PITH_FULL_IMAGE:figures/full_fig_p076_3_2.png]
Figure 3.3
Figure 3.3. Figure 3.3: The effect of applying downsampling from the intensity perspective. [PITH_FULL_IMAGE:figures/full_fig_p078_3_3.png]
Figure 3.4
Figure 3.4. Figure 3.4: A diagram to illustrate the design of a typical square fiducial marker [ [PITH_FULL_IMAGE:figures/full_fig_p079_3_4.png]
Figure 3.5
Figure 3.5. Figure 3.5: The effect of clustering on the downsampling result. [PITH_FULL_IMAGE:figures/full_fig_p080_3_5.png]
Figure 3.6
Figure 3.6. Figure 3.6: A diagram of the possible OBB size for a given marker. [PITH_FULL_IMAGE:figures/full_fig_p081_3_6.png]
Figure 3.7
Figure 3.7. Figure 3.7: The effect of filtering out the unwanted clusters. [PITH_FULL_IMAGE:figures/full_fig_p083_3_7.png]
Figure 3.8
Figure 3.8. Figure 3.8: The result of extracting points falling into the preserved OBBs from the raw [PITH_FULL_IMAGE:figures/full_fig_p085_3_8.png]
Figure 3.9
Figure 3.9. Figure 3.9: An illustration of how the intermediate plane helps solve the occlusion issue. [PITH_FULL_IMAGE:figures/full_fig_p086_3_9.png]
Figure 3.10
Figure 3.10. Figure 3.10: An illustration of how the intermediate plane helps solve the occlusion [PITH_FULL_IMAGE:figures/full_fig_p086_3_10.png]
Figure 3.11
Figure 3.11. Figure 3.11: The fiducial marker localization result after applying the intermediate plane [PITH_FULL_IMAGE:figures/full_fig_p088_3_11.png]
Figure 3.12
Figure 3.12. Figure 3.12: An illustration of the potential rotational ambiguity issue of [PITH_FULL_IMAGE:figures/full_fig_p089_3_12.png]
Figure 3.13
Figure 3.13. Figure 3.13: The LiDAR scans the indoor parking lot from right to left, moving rapidly [PITH_FULL_IMAGE:figures/full_fig_p093_3_13.png]
Figure 3.14
Figure 3.14. Figure 3.14: Traj LO [10] is utilized to create this 3D map. 67 [PITH_FULL_IMAGE:figures/full_fig_p094_3_14.png]
Figure 3.15
Figure 3.15. Figure 3.15: Comparison of the intensity images generated by the vanilla IFM and the [PITH_FULL_IMAGE:figures/full_fig_p095_3_15.png]
Figure 4.1
Figure 4.1. Figure 4.1: An overview of the proposed framework for mapping and localization using [PITH_FULL_IMAGE:figures/full_fig_p100_4_1.png]
Figure 4.2
Figure 4.2. Figure 4.2: The raw intensity image binarized with different threshold values. [PITH_FULL_IMAGE:figures/full_fig_p101_4_2.png]
Figure 4.3
Figure 4.3. Figure 4.3: An illustration of the first-level graph. After applying the proposed adaptive [PITH_FULL_IMAGE:figures/full_fig_p106_4_3.png]
Figure 4.4
Figure 4.4. Figure 4.4: The procedures for formulating the second-level graph. The variable nodes [PITH_FULL_IMAGE:figures/full_fig_p109_4_4.png]
Figure 4.5
Figure 4.5. Figure 4.5: Comparison of 3DMatch and Livox-3DMatch. (a): A random sample from [PITH_FULL_IMAGE:figures/full_fig_p112_4_5.png]
Figure 4.6
Figure 4.6. Figure 4.6: Setup for testing the adaptive threshold marker detection algorithm: (a) be [PITH_FULL_IMAGE:figures/full_fig_p114_4_6.png]
Figure 4
Figure 4. Figure 4: (b) I Fig. 4.6(c) In a large 0 7 17 0 7 9 0 7 19/20 [PITH_FULL_IMAGE:figures/full_fig_p115_4.png]
Figure 4.7
Figure 4.7. Figure 4.7: A comparison with SOTA methods, including MDGD [ [PITH_FULL_IMAGE:figures/full_fig_p118_4_7.png]
Figure 4.8
Figure 4.8. Figure 4.8: An illustration of the experimental setup and a visual comparison of the [PITH_FULL_IMAGE:figures/full_fig_p123_4_8.png]
Figure 4.9
Figure 4.9. Figure 4.9: Visual comparison of the instance reconstruction. From top to bottom: [PITH_FULL_IMAGE:figures/full_fig_p124_4_9.png]
Figure 4.10
Figure 4.10. Figure 4.10: An illustration of the scenes in Livox-3DMatch. [PITH_FULL_IMAGE:figures/full_fig_p125_4_10.png]
Figure 4.11
Figure 4.11. Figure 4.11: An illustration of the experimental setup and a visual comparison of the [PITH_FULL_IMAGE:figures/full_fig_p128_4_11.png]
Figure 4.12
Figure 4.12. Figure 4.12: A comparison of the sensor trajectories obtained from the proposed method [PITH_FULL_IMAGE:figures/full_fig_p129_4_12.png]
Figure 4.13
Figure 4.13. Figure 4.13: (a): An illustration of the experimental setup. (b): Comparison of the [PITH_FULL_IMAGE:figures/full_fig_p131_4_13.png]
Figure 4.14
Figure 4.14. Figure 4.14: A comparison of the 3D map merging results of MDGD [ [PITH_FULL_IMAGE:figures/full_fig_p133_4_14.png]
Figure 4.15
Figure 4.15. Figure 4.15: A comparison of the point cloud registration results without the first graph, [PITH_FULL_IMAGE:figures/full_fig_p135_4_15.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

96 extracted references · 76 canonical work pages

  1. [39]

    Intensity image-based lidar fiducial marker system,

    Y. Liu, H. Schofield, and J. Shan, “Intensity image-based lidar fiducial marker system,” IEEE Robotics and Automation Letters, vol. 7, no. 3, pp. 6542–6549, 2022

    work page 2022

  2. [1]

    Flexible layouts for fiducial tags,

    M. Krogius, A. Haggenmiller, and E. Olson, “Flexible layouts for fiducial tags,” in Proc. IEEE/RSJ International Conference on Intelligent Robots and Systems, 2019, pp. 1898–1903

    2019

  3. [2]

    Matching distance and geometric dis- tribution aided learning multiview point cloud registration,

    S. Li, J. Zhu, Y. Xie, N. Hu, and D. Wang, “Matching distance and geometric dis- tribution aided learning multiview point cloud registration,”IEEE Robotics and Au- tomation Letters, vol. 9, no. 11, pp. 9319–9326, 2024

    2024

  4. [3]

    Robust multiview point cloud registration with reliable pose graph initialization and history reweighting,

    H. Wang, Y. Liu, Z. Dong, Y. Guo, Y.-S. Liu, W. Wang, and B. Yang, “Robust multiview point cloud registration with reliable pose graph initialization and history reweighting,” inProc. of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023, pp. 9506–9515

    2023

  5. [4]

    Se3et: Se(3)-equivariant transformer for low- overlap point cloud registration,

    C. E. Lin, M. Zhu, and M. Ghaffari, “Se3et: Se(3)-equivariant transformer for low- overlap point cloud registration,” IEEE Robotics and Automation Letters, vol. 9, no. 11, pp. 9526–9533, 2024

    2024

  6. [5]

    Geotransformer: Fast and robust point cloud registration with geometric transformer,

    Z. Qin, H. Yu, C. Wang, Y. Guo, Y. Peng, S. Ilic, D. Hu, and K. Xu, “Geotransformer: Fast and robust point cloud registration with geometric transformer,”IEEE Trans- actions on Pattern Analysis and Machine Intelligence, vol. 45, no. 8, pp. 9806–9821, 2023. 115

    2023

  7. [6]

    Teaser: Fast and certifiable point cloud registration,

    H. Yang, J. Shi, and L. Carlone, “Teaser: Fast and certifiable point cloud registration,” IEEE Transactions on Robotics, vol. 37, no. 2, pp. 314–333, 2020

    2020

  8. [7]

    An efficient visual sfm framework using planar markers,

    Z. Jia, Y. Rao, H. Fan, and J. Dong, “An efficient visual sfm framework using planar markers,” IEEE Transactions on Instrumentation and Measurement, vol. 72, pp. 1–12, 2023

    2023

Show all 96 references
  1. [8]

    KISS- ICP: In Defense of Point-to-Point ICP – Simple, Accurate, and Robust Registration If Done the Right Way,

    I. Vizzo, T. Guadagnino, B. Mersch, L. Wiesmann, J. Behley, and C. Stachniss, “KISS- ICP: In Defense of Point-to-Point ICP – Simple, Accurate, and Robust Registration If Done the Right Way,”IEEE Robotics and Automation Letters, vol. 8, no. 2, pp. 1029–1036, 2023

    2023

  2. [9]

    Lidartag: A real-time fiducial tag system for point clouds,

    J.-K. Huang, S. Wang, M. Ghaffari, and J. W. Grizzle, “Lidartag: A real-time fiducial tag system for point clouds,”IEEE Robotics and Automation Letters, vol. 6, no. 3, pp. 4875–4882, 2021

    2021

  3. [10]

    Traj-lo: In defense of lidar-only odometry using an effective continuous-time trajectory,

    X. Zheng and J. Zhu, “Traj-lo: In defense of lidar-only odometry using an effective continuous-time trajectory,”IEEE Robotics and Automation Letters, vol. 9, no. 2, pp. 1961–1968, 2024

    1961

  4. [11]

    A note on two problems in connexion with graphs,

    E. W. Dijkstra, “A note on two problems in connexion with graphs,” inEdsger Wybe Dijkstra: His Life, Work, and Legacy, 2022, pp. 287–290

    2022

  5. [12]

    Livox Mapping

    Livox-SDK. Livox Mapping. (2020). [Online]. Available: https://github.com/ Livox-SDK/livox_mapping

    2020

  6. [13]

    Loam livox: A fast, robust, high-precision lidar odometry and mapping package for lidars of small fov,

    J. Lin and F. Zhang, “Loam livox: A fast, robust, high-precision lidar odometry and mapping package for lidars of small fov,” inProc. IEEE International Conference on Robotics and Automation, 2020, pp. 3126–3131. 116

    2020

  7. [14]

    Loam: Lidar odometry and mapping in real-time

    J. Zhang and S. Singh, “Loam: Lidar odometry and mapping in real-time.” inRobotics: Science and systems, vol. 2, no. 9. Berkeley, CA, 2014, pp. 1–9

    2014

  8. [15]

    Rangedet: In defense of range view for lidar-based 3d object detection,

    L. Fan, X. Xiong, F. Wang, N. Wang, and Z. Zhang, “Rangedet: In defense of range view for lidar-based 3d object detection,” inProc. IEEE/CVF International Confer- ence on Computer Vision, 2021

    2021

  9. [16]

    Apriltag 2: Efficient and robust fiducial detection,

    J. Wang and E. Olson, “Apriltag 2: Efficient and robust fiducial detection,” inProc. IEEE/RSJ International Conference on Intelligent Robots and Systems, 2016, pp. 4193–4198

    2016

  10. [17]

    Apriltag: A robust and flexible visual fiducial system,

    E. Olson, “Apriltag: A robust and flexible visual fiducial system,” in Proc. IEEE International Conference on Robotics and Automation, 2011, pp. 3400–3407

    2011

  11. [18]

    Speeded up de- tection of squared fiducial markers,

    F. J. Romero-Ramirez, R. Muñoz-Salinas, and R. Medina-Carnicer, “Speeded up de- tection of squared fiducial markers,”Image and vision Computing, vol. 76, pp. 38–47, 2018

    2018

  12. [19]

    Detection and Accurate Lo- calization of Circular Fiducials under Highly Challenging Conditions,

    L. Calvet, P. Gurdjos, C. Griwodz, and S. Gasparini, “Detection and Accurate Lo- calization of Circular Fiducials under Highly Challenging Conditions,” inProc. IEEE Conference on Computer Vision and Pattern Recognition, 2016, pp. 562 – 570

    2016

  13. [20]

    A practical frame- work for the development of augmented reality applications by using aruco markers,

    D. Avola, L. Cinque, G. L. Foresti, C. Mercuri, and D. Pannone, “A practical frame- work for the development of augmented reality applications by using aruco markers,” in Proc. International Conference on Pattern Recognition Applications and Methods, vol. 2, 2016, pp. 645–654

    2016

  14. [21]

    Navigation of a self-driving vehicle using one fiducial marker,

    Y. Liu, H. Schofield, and J. Shan, “Navigation of a self-driving vehicle using one fiducial marker,” inProc. IEEE International Conference on Multisensor Fusion and Integration for Intelligent Systems, 2021, pp. 1–6. 117

    2021

  15. [22]

    Application of ghost-deblurgan to fiducial marker detection,

    Y. Liu, A. Haridevan, H. Schofield, and J. Shan, “Application of ghost-deblurgan to fiducial marker detection,” inProc. IEEE/RSJ International Conference on Intelligent Robots and Systems, 2022, pp. 6827–6832

    2022

  16. [23]

    Autonomous vision-based uav landing with collision avoidance using deep learning,

    T. Liao, A. Haridevan, Y. Liu, and J. Shan, “Autonomous vision-based uav landing with collision avoidance using deep learning,” inScience and Information Conference. Springer, 2022, pp. 79–87

    2022

  17. [24]

    Extending kalibr: Calibrating the extrinsics of multiple imus and of individual axes,

    J. Rehder, J. Nikolic, T. Schneider, T. Hinzmann, and R. Siegwart, “Extending kalibr: Calibrating the extrinsics of multiple imus and of individual axes,” inProc. IEEE International Conference on Robotics and Automation, 2016, pp. 4304–4311

    2016

  18. [25]

    Improvements to target-based 3d lidar to camera calibration,

    J.-K. Huang and J. W. Grizzle, “Improvements to target-based 3d lidar to camera calibration,” IEEE Access, vol. 8, pp. 134101–134110, 2020

    2020

  19. [27]

    Spm-slam: Simul- taneous localization and mapping with squared planar markers,

    R. Muñoz-Salinas, M. J. Marin-Jimenez, and R. Medina-Carnicer, “Spm-slam: Simul- taneous localization and mapping with squared planar markers,”Pattern Recognition, vol. 86, pp. 156–171, 2019

    2019

  20. [28]

    Variational bayesian estimator for mobile robot local- ization with unknown noise covariance,

    S. Zhang, J. Shan, and Y. Liu, “Variational bayesian estimator for mobile robot local- ization with unknown noise covariance,”IEEE/ASME Transactions on Mechatronics, vol. 27, no. 4, pp. 2185–2193, 2022

    2022

  21. [29]

    Particle filtering on lie group for mobile robot localization with range-bearing measurements,

    ——, “Particle filtering on lie group for mobile robot localization with range-bearing measurements,” IEEE Control Systems Letters, vol. 7, pp. 3753–3758, 2023. 118

    2023

  22. [30]

    Multiway point cloud mosaicking with diffusion and global optimization,

    S. Jin, I. Armeni, M. Pollefeys, and D. Barath, “Multiway point cloud mosaicking with diffusion and global optimization,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024, pp. 20838–20849

    2024

  23. [31]

    Approximate inference particle filtering for mobile robot slam,

    S. Zhang, J. Shan, and Y. Liu, “Approximate inference particle filtering for mobile robot slam,” IEEE Transactions on Automation Science and Engineering, pp. 1–12, 2024

    2024

  24. [32]

    Reconstructing objects in-the-wild for realistic sensor simulation,

    Z. Yang, S. Manivasagam, Y. Chen, J. Wang, R. Hu, and R. Urtasun, “Reconstructing objects in-the-wild for realistic sensor simulation,” in Proc. of IEEE International Conference on Robotics and Automation, 2023, pp. 11661–11668

    2023

  25. [33]

    Scalability in perception for autonomous driving: Waymo open dataset,

    P. Sun, H. Kretzschmar, X. Dotiwalla, A. Chouard, V. Patnaik, P. Tsui, J. Guo, Y. Zhou, Y. Chai, B. Caineet al., “Scalability in perception for autonomous driving: Waymo open dataset,” inProc. of the IEEE/CVF conference on computer vision and pattern recognition, 2020, pp. 2446–2454

    2020

  26. [34]

    Mv-deepsdf: Implicit modeling with multi-sweep point clouds for 3d vehicle reconstruction in autonomous driving,

    Y. Liu, K. Zhu, G. Wu, Y. Ren, B. Liu, Y. Liu, and J. Shan, “Mv-deepsdf: Implicit modeling with multi-sweep point clouds for 3d vehicle reconstruction in autonomous driving,” in Proc. of the IEEE/CVF International Conference on Computer Vision, 2023, pp. 8306–8316

    2023

  27. [35]

    3dmatch: Learning local geometric descriptors from rgb-d reconstructions,

    A. Zeng, S. Song, M. Nießner, M. Fisher, J. Xiao, and T. Funkhouser, “3dmatch: Learning local geometric descriptors from rgb-d reconstructions,” inProc. of the IEEE conference on computer vision and pattern recognition, 2017, pp. 1802–1811

    2017

  28. [36]

    Challenging data sets for point cloudregistrationalgorithms,

    F. Pomerleau, M. Liu, F. Colas, and R. Siegwart, “Challenging data sets for point cloudregistrationalgorithms,” The International Journal of Robotics Research, vol.31, no. 14, pp. 1705–1711, 2012. 119

    2012

  29. [37]

    Scannet: Richly-annotated 3d reconstructions of indoor scenes,

    A. Dai, A. X. Chang, M. Savva, M. Halber, T. Funkhouser, and M. Nießner, “Scannet: Richly-annotated 3d reconstructions of indoor scenes,” in Proceedings of the IEEE conference on computer vision and pattern recognition, 2017, pp. 5828–5839

    2017

  30. [38]

    Mending neural implicit modeling for 3d vehicle reconstruction in the wild,

    S. Duggal, Z. Wang, W.-C. Ma, S. Manivasagam, J. Liang, S. Wang, and R. Urtasun, “Mending neural implicit modeling for 3d vehicle reconstruction in the wild,” inProc. of the IEEE/CVF Winter Conference on Applications of Computer Vision, 2022, pp. 1900–1909

    2022

  31. [40]

    Mapping and localization from planar markers,

    R. Muñoz-Salinas, M. J. Marín-Jimenez, E. Yeguas-Bolivar, and R. Medina-Carnicer, “Mapping and localization from planar markers,” Pattern Recognition, vol. 73, pp. 158–171, 2018

    2018

  32. [41]

    Automatic extrinsic cali- bration method for lidar and camera sensor setups,

    J. Beltrán, C. Guindel, A. de la Escalera, and F. García, “Automatic extrinsic cali- bration method for lidar and camera sensor setups,”IEEE Transactions on Intelligent Transportation Systems, vol. 23, no. 10, pp. 17677–17689, 2022

    2022

  33. [42]

    Cpi: Lidar-camera extrinsic calibration based on feature points with reflection intensity,

    L. Tao, L. Pei, T. Li, D. Zou, Q. Wu, and S. Xia, “Cpi: Lidar-camera extrinsic calibration based on feature points with reflection intensity,” in Spatial Data and Intelligence: First International Conference, SpatialDI 2020, Virtual Event, May 8–9, 2020, Proceedings 1. Springe...

    2020

  34. [43]

    A4lidartag: Depth-based fiducial marker for extrinsic calibration of solid-state lidar and camera,

    Y. Xie, L. Deng, T. Sun, Y. Fu, J. Li, X. Cui, H. Yin, S. Deng, J. Xiao, and B. Chen, “A4lidartag: Depth-based fiducial marker for extrinsic calibration of solid-state lidar and camera,” IEEE Robotics and Automation Letters, vol. 7, no. 3, pp. 6487–6494, 2022. 120

    2022

  35. [44]

    Random sample consensus: a paradigm for model fitting with applications to image analysis and automated cartography,

    M. A. Fischler and R. C. Bolles, “Random sample consensus: a paradigm for model fitting with applications to image analysis and automated cartography,”Communica- tions of the ACM, vol. 24, no. 6, pp. 381–395, 1981

    1981

  36. [45]

    Predator: Regis- tration of 3d point clouds with low overlap,

    S. Huang, Z. Gojcic, M. Usvyatsov, A. Wieser, and K. Schindler, “Predator: Regis- tration of 3d point clouds with low overlap,” inProc. of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, June 2021, pp. 4267–4276

    2021

  37. [46]

    Recognition of degradation scenarios for lidar slam applications,

    C. Yang, Z. Chai, X. Yang, H. Zhuang, and M. Yang, “Recognition of degradation scenarios for lidar slam applications,” in Proc. IEEE International Conference on Robotics and Biomimetics. IEEE, 2022, pp. 1726–1731

    2022

  38. [47]

    Structure-from-motion revisited,

    J. L. Schonberger and J.-M. Frahm, “Structure-from-motion revisited,” inProceedings of the IEEE conference on computer vision and pattern recognition, 2016, pp. 4104– 4113

    2016

  39. [48]

    Dust3r: Geometric 3d vision made easy,

    S. Wang, V. Leroy, Y. Cabon, B. Chidlovskii, and J. Revaud, “Dust3r: Geometric 3d vision made easy,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024, pp. 20697–20709

    2024

  40. [49]

    Omniobject3d: Large-vocabulary 3d object dataset for realistic per- ception, reconstruction and generation,

    T. Wu, J. Zhang, X. Fu, Y. Wang, J. Ren, L. Pan, W. Wu, L. Yang, J. Wang, C. Qianet al., “Omniobject3d: Large-vocabulary 3d object dataset for realistic per- ception, reconstruction and generation,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recogn...

    2023

  41. [50]

    Arc-support line segments revisited: An efficient high-quality ellipse detection,

    C. Lu, S. Xia, M. Shao, and Y. Fu, “Arc-support line segments revisited: An efficient high-quality ellipse detection,”IEEE Transactions on Image Processing, vol. 29, pp. 768–781, 2019. 121

    2019

  42. [51]

    Circle detection by arc-support line segments,

    C. Lu, S. Xia, W. Huang, M. Shao, and Y. Fu, “Circle detection by arc-support line segments,” in Proc. IEEE International Conference on Image Processing, 2017, pp. 76–80

    2017

  43. [52]

    Infinitesimal plane-based pose estimation,

    T. Collins and A. Bartoli, “Infinitesimal plane-based pose estimation,”International journal of computer vision, vol. 109, no. 3, pp. 252–286, 2014

    2014

  44. [53]

    Navigation of a self-driving vehicle using one fiducial marker,

    Y. Liu, H. Schofield, and J. Shan, “Navigation of a self-driving vehicle using one fiducial marker,” inProc. IEEE International Conference on Multisensor Fusion and Integration for Intelligent Systems, 2021, pp. 1–6

    2021

  45. [54]

    Resolving marker pose ambiguity by robust rotation averaging with clique constraints,

    S.-F. Ch’ng, N. Sogi, P. Purkait, T.-J. Chin, and K. Fukui, “Resolving marker pose ambiguity by robust rotation averaging with clique constraints,” inProc. IEEE Inter- national Conference on Robotics and Automation, 2020, pp. 9680–9686

    2020

  46. [55]

    Accurate outdoor ground truth based on total stations,

    M. Vaidis, P. Giguère, F. Pomerleau, and V. Kubelka, “Accurate outdoor ground truth based on total stations,” in2021 18th Conference on Robots and Vision (CRV), 2021, pp. 1–8

    2021

  47. [56]

    Unified temporal and spatial calibration for multi-sensor systems,

    P. Furgale, J. Rehder, and R. Siegwart, “Unified temporal and spatial calibration for multi-sensor systems,” in Porc. IEEE/RSJ International Conference on Intelligent Robots and Systems, 2013, pp. 1280–1286

    2013

  48. [57]

    Fast range image segmentation by an edge detection strategy,

    A. D. Sappa and M. Devy, “Fast range image segmentation by an edge detection strategy,” in Proc. International Conference on 3-D Digital Imaging and Modeling, 2001, pp. 292–299

    2001

  49. [58]

    Hierarchical clustering schemes,

    S. C. Johnson, “Hierarchical clustering schemes,”Psychometrika, vol. 32, no. 3, pp. 241–254, 1967. 122

    1967

  50. [59]

    F-loam: Fast lidar odometry and map- ping,

    H. Wang, C. Wang, C.-L. Chen, and L. Xie, “F-loam: Fast lidar odometry and map- ping,” in2021 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). IEEE, 2021, pp. 4390–4396

    2021

  51. [60]

    Tong and C

    Q. Tong and C. Shaozu. A-LOAM: Advanced implementation of loam. (2019). [Online]. Available: âĂİhttps://github.com/HKUST-Aerial-Robotics/A-LOAM

    2019

  52. [61]

    Learning multiview 3d point cloud registration,

    Z. Gojcic, C. Zhou, J. D. Wegner, L. J. Guibas, and T. Birdal, “Learning multiview 3d point cloud registration,” inProc. of the IEEE/CVF conference on computer vision and pattern recognition, 2020, pp. 1759–1769

    2020

  53. [62]

    Learning compact geometric features,

    M. Khoury, Q.-Y. Zhou, and V. Koltun, “Learning compact geometric features,” in Proc. of the IEEE international conference on computer vision, 2017, pp. 153–161

    2017

  54. [63]

    Improvements to thin-sheet 3d lidar fiducial tag localization,

    Y. Liu, J. Shan, and H. Schofield, “Improvements to thin-sheet 3d lidar fiducial tag localization,” IEEE Access, vol. 12, pp. 124907–124914, 2024

    2024

  55. [64]

    L-pr: Exploiting lidar fiducial marker for unordered low overlap multiview point cloud registration,

    Y. Liu, J. Shan, A. Haridevan, and S. Zhang, “L-pr: Exploiting lidar fiducial marker for unordered low overlap multiview point cloud registration,”IEEE Transactions on Instrumentation and Measurement, 2025

    2025

  56. [65]

    T. D. Barfoot,State estimation for robotics. Cambridge University Press, 2017

    2017

  57. [66]

    Tagslam: Robust slam with fiducial markers,

    B. Pfrommer and K. Daniilidis, “Tagslam: Robust slam with fiducial markers,”arXiv preprint arXiv:1910.00679, 2019

    1910 arXiv

  58. [67]

    Lasernet: An efficient probabilistic 3d object detector for autonomous driving,

    G. P. Meyer, A. Laddha, E. Kee, C. Vallespi-Gonzalez, and C. K. Wellington, “Lasernet: An efficient probabilistic 3d object detector for autonomous driving,” in Proc. IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2019, pp. 12677–12686. 123

    2019

  59. [68]

    Good colour maps: How to design them,

    P. Kovesi, “Good colour maps: How to design them,”arXiv preprint arXiv:1509.03700, 2015

    2015 arXiv

  60. [69]

    3d is here: Point cloud library (pcl),

    R. B. Rusu and S. Cousins, “3d is here: Point cloud library (pcl),” inProc. IEEE International Conference on Robotics and Automation, 2011, pp. 1–4

    2011

  61. [70]

    A class of fast gaussian binomial filters for speech and image processing,

    R. A. Haddad, A. N. Akansuet al., “A class of fast gaussian binomial filters for speech and image processing,” IEEE Transactions on Signal Processing, vol. 39, no. 3, pp. 723–727, 1991

    1991

  62. [71]

    Nicp: Dense normal based point cloud registration,

    J. Serafin and G. Grisetti, “Nicp: Dense normal based point cloud registration,” in Proc. IEEE/RSJ International Conference on Intelligent Robots and Systems, 2015, pp. 742–749

    2015

  63. [72]

    Least-squares fitting of two 3-d point sets,

    K. S. Arun, T. S. Huang, and S. D. Blostein, “Least-squares fitting of two 3-d point sets,” IEEE Transactions on pattern analysis and machine intelligence, no. 5, pp. 698–700, 1987

    1987

  64. [73]

    Scalable fiducial tag localization on a 3d prior map via graph-theoretic global tag-map registration,

    K. Koide, S. Oishi, M. Yokozuka, and A. Banno, “Scalable fiducial tag localization on a 3d prior map via graph-theoretic global tag-map registration,” inProc. of IEEE/RSJ International Conference on Intelligent Robots and Systems. IEEE, 2022, pp. 5347– 5353

    2022

  65. [74]

    Vins-mono: A robust and versatile monocular visual- inertial state estimator,

    T. Qin, P. Li, and S. Shen, “Vins-mono: A robust and versatile monocular visual- inertial state estimator,”IEEE transactions on robotics, vol. 34, no. 4, pp. 1004–1020, 2018

    2018

  66. [75]

    Semantic 3d object maps for everyday manipulation in human living environments,

    R. B. Rusu, “Semantic 3d object maps for everyday manipulation in human living environments,” KI-Künstliche Intelligenz, vol. 24, no. 4, pp. 345–348, 2010

    2010

  67. [76]

    Schneider and D

    P. Schneider and D. H. Eberly,Geometric tools for computer graphics. Elsevier, 2002. 124

    2002

  68. [77]

    Cloudcompare,

    D. Girardeau-Montautet al., “Cloudcompare,”France: EDF R&D Telecom ParisTech, vol. 11, no. 5, 2016

    2016

  69. [78]

    isam2: In- cremental smoothing and mapping with fluid relinearization and incremental variable reordering,

    M. Kaess, H. Johannsson, R. Roberts, V. Ila, J. Leonard, and F. Dellaert, “isam2: In- cremental smoothing and mapping with fluid relinearization and incremental variable reordering,” inProc. of IEEE International Conference on Robotics and Automation, 2011, pp. 3281–3288

    2011

  70. [79]

    Dellaert and M

    F. Dellaert and M. Kaess, Factor Graphs for Robot Perception. Foundations and Trends in Robotics, Vol. 6, 2017. [Online]. Available: http://www.cs.cmu.edu/ ~kaess/pub/Dellaert17fnt.pdf

    2017

  71. [80]

    Supervisely Computer Vision platform

    Supervisely. Supervisely Computer Vision platform. (2023). [Online]. Available: https://supervisely.com

    2023

  72. [81]

    Fiducial tag localization on a 3d lidar prior map,

    Y. Liu, J. Shan, and H. Schofield, “Fiducial tag localization on a 3d lidar prior map,”

  73. [82]

    Unsupervised intrinsic image decomposition with lidar intensity,

    S. Sato, Y. Yao, T. Yoshida, T. Kaneko, S. Ando, and J. Shimamura, “Unsupervised intrinsic image decomposition with lidar intensity,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023, pp. 13466–13475

    2023

  74. [83]

    Neural lidar fields for novel view synthesis,

    S. Huang, Z. Gojcic, Z. Wang, F. Williams, Y. Kasten, S. Fidler, K. Schindler, and O. Litany, “Neural lidar fields for novel view synthesis,” inProceedings of the IEEE/CVF International Conference on Computer Vision, 2023, pp. 18236–18246

    2023

  75. [84]

    Lidar4d: Dynamic neural fields for novel space-time view lidar synthesis,

    Z. Zheng, F. Lu, W. Xue, G. Chen, and C. Jiang, “Lidar4d: Dynamic neural fields for novel space-time view lidar synthesis,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024, pp. 5145–5154. 125 List of Publications

    2024

  76. [86]

    Navigation of a Self-Driving Vehicle Using One Fiducial Marker,

    Y. Liu, H. Schofield and J. Shan, "Navigation of a Self-Driving Vehicle Using One Fiducial Marker," IEEE International Conference on Multisensor Fusion and Inte- gration for Intelligent Systems (MFI), Karlsruhe, Germany, 2021, pp. 1-6, DOI: 10.1109/MFI52462.2021.9591194

    2021

  77. [87]

    Intensity Image-Based LiDAR Fiducial Marker System,

    Y. Liu, H. Schofield and J. Shan, "Intensity Image-Based LiDAR Fiducial Marker System," in IEEE Robotics and Automation Letters (RA-L), vol. 7, no. 3, pp. 6542-6549, 2022, DOI:10.1109/LRA.2022.3174971

    2022

  78. [88]

    Application of Ghost-DeblurGAN to Fiducial Marker Detection,

    Y. Liu, A. Haridevan, H. Schofield and J. Shan, "Application of Ghost-DeblurGAN to Fiducial Marker Detection," IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), Kyoto, Japan, 2022, pp. 6827-6832, DOI:10.1109/ IROS47612.2022.9981701

    2022

  79. [89]

    MV-DeepSDF: Implicit Modeling with Multi-Sweep Point Clouds for 3D Vehicle Reconstruction in Autonomous Driving,

    Y. Liu , K. Zhu, G. Wu. Y. Ren, B. Liu, Y. Liu, and J. Shan, "MV-DeepSDF: Implicit Modeling with Multi-Sweep Point Clouds for 3D Vehicle Reconstruction in Autonomous Driving," IEEE/CVF International Conference on Computer Vi- sion (ICCV), Paris, France, 2023, pp. 8272-8282, DO...

    2023

  80. [90]

    VQA-Diff: 126 ExploitingVQAandDiffusionforZero-ShotImage-to-3DVehicleAssetGenerationin Autonomous Driving,

    Y. Liu*, Z. Yang*, G. Wu. Y. Ren, K. Lin, B. Liu, Y. Liu, and J. Shan, "VQA-Diff: 126 ExploitingVQAandDiffusionforZero-ShotImage-to-3DVehicleAssetGenerationin Autonomous Driving," European Conference on Computer Vision (ECCV), Milan, Italy, 2024. pp. 323-340, DOI:https://doi.o...

    2024 doi

  81. [91]

    Improvements to Thin-Sheet 3D LiDAR Fiducial Tag Localization,

    Y. Liu, J. Shan, and H. Schofield, "Improvements to Thin-Sheet 3D LiDAR Fiducial Tag Localization,", IEEE ACCESS, vol. 12, pp. 124907-124914, 2024, DOI:10. 1109/ACCESS.2024.3451404

    2024

  82. [92]

    L-PR: Exploiting LiDAR Fidu- cial Marker for Unordered Low Overlap Multiview Point Cloud Registration,

    Y. Liu , J. Shan, A. Haridevan, and S. Zhang, “L-PR: Exploiting LiDAR Fidu- cial Marker for Unordered Low Overlap Multiview Point Cloud Registration,” IEEE Transactions on Instrumentation & Measurement (TIM), in press, 2025

    2025

  83. [93]

    Approximate Inference Particle Filtering for Mobile Robot SLAM,

    S. Zhang, J. Shan, andY. Liu, "Approximate Inference Particle Filtering for Mobile Robot SLAM," in IEEE Transactions on Automation Science and Engineering (T- ASE), pp. 1-12, 2024, DOI:10.1109/TASE.2024.3475735

    2024

  84. [94]

    Learning Effective NeRFs and SDFs Representations with 3D GANs for Object Generation,

    Z. Yang*,Y. Liu*, G. Wu, T. Cao, Y. Ren, Y. Liu, and B. Liu, “Learning Effective NeRFs and SDFs Representations with 3D GANs for Object Generation,”NeurIPS W orkshop, 2024 (* co-first author)

    2024

  85. [95]

    Variational Bayesian Estimator for Mobile Robot Localization With Unknown Noise Covariance,

    S. Zhang, J. Shan, andY. Liu, "Variational Bayesian Estimator for Mobile Robot Localization With Unknown Noise Covariance," in IEEE/ASME Transactions on Mechatronics (T-MECH), vol. 27, no. 4, pp. 2185-2193, 2022, DOI: 10.1109/ TMECH.2022.3161591

    2022

  86. [96]

    Particle Filtering on Lie Group for Mobile Robot Localization With Range-Bearing Measurements,

    S. Zhang, J. Shan, andY. Liu, "Particle Filtering on Lie Group for Mobile Robot Localization With Range-Bearing Measurements," in IEEE Control Systems Letters (L-CSS), vol. 7, pp. 3753-3758, 2023, DOI:10.1109/LCSYS.2023.3340419. 127

    2023

  87. [97]

    Autonomous Vision-Based UAV Land- ing with Collision Avoidance Using Deep Learning,

    T. Liao, A. Haridevan,Y. Liu, and J. Shan, “Autonomous Vision-Based UAV Land- ing with Collision Avoidance Using Deep Learning,” Science and Information Con- ference (SAI), pp. 79–87, 2022, DOI:10.1007/978-3-031-10464-0_6 128

    2022 doi

  88. [2024]

    Available: https://arxiv.org/abs/2209.01072

    [Online]. Available: https://arxiv.org/abs/2209.01072

    arXiv

Pith tools

Reviewed August 9, 2026 · model on record in the stance chip above.