Keypoint Matching for Point Cloud Registration Using Multiplex Dynamic Graph Attention Networks

Shi, Chenghao; Chen, Xieyuanli; Huang, Kaixun; Xiao, Jing; Lu, Huimin; Stachniss, Cyrill

doi:10.1109/lra.2021.3097275

Cited by 50 publications

(34 citation statements)

References 32 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The SpinNet [35] is a new neural architecture which projects the point cloud into a cylindrical space; then the descriptor is invariably extracted rotationally by a special convolutional neural layer. Inspired by the success of GAT for computer vision tasks, Shi et al [4] proposed a multiplex dynamic graph attention network matching method (MDGAT-matcher). It uses a novel and flexible graph network architecture on the point cloud and achieves the enriched feature representation by recovering local information.…”

Section: Keypoint Detectormentioning

confidence: 99%

“…To this end, we find inspiration in the MDGAT-matcher [4] for a GAT-based two-stage descriptor for point cloud registration. However, the extension of our work is non-trivial.…”

Section: Keypoint Detectormentioning

confidence: 99%

“…In the MDGAT-matcher [4], a dynamic GAT was introduced to embed the geometry feature from the FPFH descriptor to create more robust feature for the matching process. Because of the limitations of handcrafted descriptors, their results are sensitive and rotationvariant to rigid transformation in Euclidian space.…”

Section: Graph Attention Networkmentioning

confidence: 99%

“…Efficient keypoint matching is required for different applications such as 3D object identification, simultaneous localization and mapping (SLAM), sparse depth odometry, 3D shape retrieval, and range image/point cloud registration [1][2][3][4]. All of these techniques include keypoint recognition and their matching by employing element descriptors to find genuine keypoint correspondences.…”

Section: Introductionmentioning

confidence: 99%

“…detect keypoints from 3D point clouds, but also derive compact and representative descriptors for matching. To this end, we find inspiration in the MDGAT-matcher [4] for a GAT-based twostage descriptor for point cloud registration. However, the extension of our work is nontrivial.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Learning Dense Features for Point Cloud Registration Using a Graph Attention Network

2022

View full text Add to dashboard Cite

Point cloud registration is a fundamental task in many applications such as localization, mapping, tracking, and reconstruction. Successful registration relies on extracting robust and discriminative geometric features. Though existing learning-based methods require high computing capacity for processing a large number of raw points at the same time, computational capacity limitation is not an issue thanks to powerful parallel computing process using GPU. In this paper, we introduce a framework that efficiently and economically extracts dense features using a graph attention network for point cloud matching and registration (DFGAT). The detector of the DFGAT is responsible for finding highly reliable key points in large raw data sets. The descriptor of the DFGAT takes these keypoints combined with their neighbors to extract invariant density features in preparation for the matching. The graph attention network (GAT) uses the attention mechanism that enriches the relationships between point clouds. Finally, we consider this as an optimal transport problem and use the Sinkhorn algorithm to find positive and negative matches. We perform thorough tests on the KITTI dataset and evaluate the effectiveness of this approach. The results show that this method with the efficiently compact keypoint selection and description can achieve the best performance matching metrics and reach the highest success ratio of 99.88% registration in comparison with other state-of-the-art approaches.

show abstract

Section: Keypoint Detectormentioning

confidence: 99%

“…To this end, we find inspiration in the MDGAT-matcher [4] for a GAT-based two-stage descriptor for point cloud registration. However, the extension of our work is non-trivial.…”

Section: Keypoint Detectormentioning

confidence: 99%

Section: Graph Attention Networkmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Learning Dense Features for Point Cloud Registration Using a Graph Attention Network

2022

View full text Add to dashboard Cite

show abstract

LoGDesc: Local Geometric Features Aggregation for Robust Point Cloud Registration

Slimani,

Tamadazte,

Achard

2024

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Rocnet: 3D robust registration of points clouds using deep learning

Slimani,

Tamadazte,

Achard

2024

Machine Vision and Applications

View full text Add to dashboard Cite

This paper introduces a new method for 3D point cloud registration based on deep learning. The architecture is composed of three distinct blocs: (i) an encoder with a convolutional graph-based descriptor that encodes the immediate neighbourhood of each point and an attention mechanism that encodes the variations of the surface normals. Such descriptors are refined by highlighting attention between the points of the same set (source and target) and then between the points of the two sets. (ii) a matching process that estimates a matrix of correspondences using the Sinkhorn algorithm. (iii) Finally, the rigid transformation between the two point clouds is calculated by RANSAC using the best scores of the correspondence matrix. We conduct experiments on the ModelNet40 dataset and real-world Bunny dataset, and our proposed architecture shows very promising results, outperforming state-of-the-art methods in most of the simulated configurations.

show abstract

Keypoint Matching for Point Cloud Registration Using Multiplex Dynamic Graph Attention Networks

Cited by 50 publications

References 32 publications

Learning Dense Features for Point Cloud Registration Using a Graph Attention Network

Learning Dense Features for Point Cloud Registration Using a Graph Attention Network

LoGDesc: Local Geometric Features Aggregation for Robust Point Cloud Registration

Rocnet: 3D robust registration of points clouds using deep learning

Contact Info

Product

Resources

About