Self-Supervised Point Cloud Registration With Deep Versatile Descriptors for Intelligent Driving

Liu, Dongrui; Chen, Chuanchuan; Xu, Changqing; Qiu, Robert C.; Chu, Lei

doi:10.1109/tits.2023.3268273

Cited by 4 publications

(2 citation statements)

References 61 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…3DSmoothNet 24 proposes a rotation-invariant 3D descriptor. Reference 25 is recommended to use both global and local descriptors to generate descriptors in a self-supervised manner. More keypoint descriptions can be found in Ref.…”

Section: Related Workmentioning

confidence: 99%

Two-stage outlier removal strategy for correspondence-based point cloud registration

Li,

Chen,

Gao

2023

J. Appl. Rem. Sens.

View full text Add to dashboard Cite

.With the emergence of keypoint matching technology, the correspondence-based point cloud registration (PCR) method has gained increasing attention. However, the correspondences generated by keypoint matching technology contain extremely high outlier rates, resulting in the correspondence-based PCR method facing issues of high computational complexity and low precision registration. We propose a correspondence-based PCR method using a coarse-to-fine outlier removal strategy with O ( N ) complexity. First, we propose a coarse outlier removal module based on linearly related properties, i.e., we build a deviation matrix that can measure each correspondence deviation the degree away from the ideal inlier. The module can reduce the number of correspondences and the outlier rates. Then, we propose a fine outlier removal module that adopts each correspondence to identify outlier based on the spatial geometric mapping invariance. Finally, to increase registration accuracy, we introduce a graduated non-convexity with Tukey’s biweight method. It can avoid the solution falling into the local minimum and better reduce the influence of outliers. Experimental results show that the proposed method is robust at outlier rates above 99% and is faster than state-of-the-art methods.

show abstract

Section: Related Workmentioning

confidence: 99%

Two-stage outlier removal strategy for correspondence-based point cloud registration

Li,

Chen,

Gao

2023

J. Appl. Rem. Sens.

View full text Add to dashboard Cite

show abstract

“…and KPConv [74] are further proposed to enhance industrial applications based on the point cloud, including recognition [38], [41], [62], detection [19], [91], registration [34], [39], [66], [92], sampling [5], [30], [37], generation [6], [46], and interpretation [69].…”

Section: D Data Processingmentioning

confidence: 99%

Wavelet-based self-supervised learning for multi-scene image fusion

Liu

Qiao

et al. 2022

Neural Comput & Applic

View full text Add to dashboard Cite

Neural Scene Flow Prior (NSFP) and Fast Neural Scene Flow (FNSF) have shown remarkable adaptability in the context of large out-of-distribution autonomous driving. Despite their success, the underlying reasons for their astonishing generalization capabilities remain unclear. Our research addresses this gap by examining the generalization capabilities of NSFP through the lens of uniform stability, revealing that its performance is inversely proportional to the number of input point clouds. This finding sheds light on NSFP's effectiveness in handling large-scale point cloud scene flow estimation tasks. Motivated by such theoretical insights, we further explore the improvement of scene flow estimation by leveraging historical point clouds across multiple frames, which inherently increases the number of point clouds. Consequently, we propose a simple and effective method for multi-frame point cloud scene flow estimation, along with a theoretical evaluation of its generalization abilities. Our analysis confirms that the proposed method maintains a limited generalization error, suggesting that adding multiple frames to the scene flow optimization process does not detract from its generalizability. Extensive experimental results on large-scale autonomous driving Waymo Open and Argoverse lidar datasets demonstrate that the proposed method achieves state-of-the-art performance.

show abstract