On Bundle Adjustment for Multiview Point Cloud Registration

Huang, Huaiyang; Sun, Yuxiang; Wu, Jin; Jiao, Jianhao; Hu, Xiangcheng; Zheng, Linwei; Wang, Lujia; Liu, Ming

doi:10.1109/lra.2021.3105686

Cited by 19 publications

(15 citation statements)

References 27 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Consequently, the method is hard to be used in large-scale problems where the lidar points are huge in number. This problem is partially addressed in [54], which aggregates all points associated to the same plane feature in a scan in the scan local frame. However, to ensure convergence, [54] modifies the cost function by including an extra heuristic penalty term, which is not a true representation of the map consistency.…”

Section: B Bundle or Plane Adjustmentmentioning

confidence: 99%

“…This problem is partially addressed in [54], which aggregates all points associated to the same plane feature in a scan in the scan local frame. However, to ensure convergence, [54] modifies the cost function by including an extra heuristic penalty term, which is not a true representation of the map consistency. Moreover, the cost function in [54] still involves the plane feature similar to [47,49]- [51,53].…”

Section: B Bundle or Plane Adjustmentmentioning

confidence: 99%

“…However, to ensure convergence, [54] modifies the cost function by including an extra heuristic penalty term, which is not a true representation of the map consistency. Moreover, the cost function in [54] still involves the plane feature similar to [47,49]- [51,53]. To lower the computation load caused by optimizing the large number of feature parameters, the method further fixes the feature parameters in the optimisation, which could slow down the optimisation speed.…”

Section: B Bundle or Plane Adjustmentmentioning

confidence: 99%

“…Consequently, the computational complexity is irrelevant to both the feature dimension (similar to BALM [32]) and the point number (similar to [47,49]). The point cluster in our method is similar to the point aggregation used in [54], but the overall BA formulation is fundamentally different: (1) it minimizes the true map consistency (the point-to-plane distance) without trading off with any other heuristic penalty; and (2) it performs exact feature elimination with rigorous proof instead of empirical fixation. Based on these nice theoretical results, we develop an efficient second-order solver, termed as BALM2.0.…”

Section: B Bundle or Plane Adjustmentmentioning

confidence: 99%

“…To verify the effectiveness of our method, we compare it with four state-of-the-art methods that focus on the lidar bundle adjustment (or similar) problem: Eigen-Factor (EF) [53], BALM [32], Plane Adjustment (PA) [49] and BAREG [54]. Among them, EF 3 , BALM 4 , BAREG 5 are open sourced, so we use the available implementation on Github.…”

Section: Benchmark Evaluationmentioning

confidence: 99%

See 4 more Smart Citations

Efficient and Consistent Bundle Adjustment on Lidar Point Clouds

Liu¹,

Liu²,

Zhang³

2022

Preprint

View full text Add to dashboard Cite

Bundle Adjustment (BA) refers to the problem of simultaneous determination of sensor poses and scene geometry, which is a fundamental problem in robot vision. This paper presents an efficient and consistent bundle adjustment method for lidar sensors. The method employs edge and plane features to represent the scene geometry, and directly minimizes the natural Euclidean distance from each raw point to the respective geometry feature. A nice property of this formulation is that the geometry features can be analytically solved, drastically reducing the dimension of the numerical optimization. To represent and solve the resultant optimization problem more efficiently, this paper then proposes a novel concept point clusters, which encodes all raw points associated to the same feature by a compact set of parameters, the point cluster coordinates. We derive the closed-form derivatives, up to the second order, of the BA optimization based on the point cluster coordinates and show their theoretical properties such as the null spaces and sparsity. Based on these theoretical results, this paper develops an efficient second-order BA solver. Besides estimating the lidar poses, the solver also exploits the second order information to estimate the pose uncertainty caused by measurement noises, leading to consistent estimates of lidar poses. Moreover, thanks to the use of point cluster, the developed solver fundamentally avoids the enumeration of each raw point (which is very time-consuming due to the large number) in all steps of the optimization: cost evaluation, derivatives evaluation and uncertainty evaluation.The proposed method is extensively evaluated at different levels: consistency, accuracy, and computation efficiency in both simulated and actual environments. Benchmark evaluation on 19 real-world open sequences covering various datasets (Hilti, NTU-VIRAL and UrbanLoco), environments (campus, urban streets, offices, laboratory, and construction sites), lidar types (Ouster OS0-64, Ouster OS1-16, Velodyne HDL 32E), and motion types (handheld, UAV-based, and ground vehicles-based) shows that our method achieves consistently and significantly higher performance than other state-of-the-art counterparts in terms of localization accuracy, mapping quality, and computation efficiency. In particular, our method achieves a mapping accuracy at a level of the lidar measurement noise (i.e., a few centimeters) while processing all sequences in less than half minute on a standard desktop CPU. Finally, we show how our proposed method effectively improves the accuracy and/or computation efficiency of some important robotic techniques, including lidarinertial odometry, multi-lidar extrinsic calibration, and highaccuracy global mapping. The implementation of our method is open sourced to benefit the robotics community and beyond 2 .

show abstract

Section: B Bundle or Plane Adjustmentmentioning

confidence: 99%

Section: B Bundle or Plane Adjustmentmentioning

confidence: 99%

Section: B Bundle or Plane Adjustmentmentioning

confidence: 99%

Section: B Bundle or Plane Adjustmentmentioning

confidence: 99%

Section: Benchmark Evaluationmentioning

confidence: 99%

See 3 more Smart Citations

Efficient and Consistent Bundle Adjustment on Lidar Point Clouds

Liu¹,

Liu²,

Zhang³

2022

Preprint

View full text Add to dashboard Cite

show abstract

MULO: LiDAR Odometry via MUlti-Landmarks Joint Optimization

Liu,

He,

Zhao

et al. 2024

J Intell Robot Syst

View full text Add to dashboard Cite

Rockfall feature investigation and kinematic simulation based on nap-of-the-object photogrammetry and GIS spatial modeling

Fuzhen

Zhao

et al. 2023

Nat Hazards

View full text Add to dashboard Cite

Chaya County, as one of the important towns in the "Three Rivers Region" in southeastern Tibet, is the key target for future expansion and development. Disaster risk investigation and research are the primary task of county expansion and development. UAVs have been widely used in disaster surveys by virtue of their ability to rapidly acquire rich and diverse geospatial data. However, due to the comprehensive in uence of terrain environment, GPS positioning accuracy, and image acquisition and processing methods, the positioning accuracy and resolution acquired from the 3D model and DSM data are di cult to guarantee its reliability. In this case, when interpreting rock mass characteristics, there will be limitations such as inaccurate extraction of joint and ssure information, inability to re ect rock mass surface roughness, and even incapacity to identify ne joints. Meanwhile, the resolution of terrain data will also affect the reliability of the simulation results of rockfall motion characteristics. Therefore, a new image acquisition method and modeling idea are summarized in this paper, and obtains a real-world 3D model with a positioning accuracy of less than 5 cm and DSM data with sub-centimeter resolution for a slope in the study area by this way. According to these data, a rock mass characteristic survey was completed, and a 3D numerical simulation of the potential rockfall movement characteristics was performed in GIS. At present, there are three potential rockfall areas in the study area, with a volume of about 2690m³. Affected by factors such as freeze-thaw erosion, differential weathering, faults, and unfavorable structural plane combinations, rockfall disasters may continue to occur in the future. The simulation results show that after the high-level rockfall occurs, the movement trajectory presents a diversion phenomenon, and the side of the township road is more threatened than the side of the G349 National Highway. The results of this paper have important reference value for disaster prevention and mitigation for the expansion of Chaya County, and can also provide important basic data for the research on fault activity, landform formation and evolution in Changdu and surrounding areas.

show abstract

On Bundle Adjustment for Multiview Point Cloud Registration

Cited by 19 publications

References 27 publications

Efficient and Consistent Bundle Adjustment on Lidar Point Clouds

Efficient and Consistent Bundle Adjustment on Lidar Point Clouds

MULO: LiDAR Odometry via MUlti-Landmarks Joint Optimization

Rockfall feature investigation and kinematic simulation based on nap-of-the-object photogrammetry and GIS spatial modeling

Contact Info

Product

Resources

About