TY - GEN
T1 - 3D Correspondence Grouping with Compatibility Features
AU - Yang, Jiaqi
AU - Chen, Jiahao
AU - Huang, Zhiqiang
AU - Cao, Zhiguo
AU - Zhang, Yanning
N1 - Publisher Copyright:
© 2021, Springer Nature Switzerland AG.
PY - 2021
Y1 - 2021
N2 - We present a simple yet effective method for 3D correspondence grouping. The objective is to accurately classify initial correspondences obtained by matching local geometric descriptors into inliers and outliers. Although the spatial distribution of correspondences is irregular, inliers are expected to be geometrically compatible with each other. Based on such observation, we propose a novel feature representation for 3D correspondences, dubbed compatibility feature (CF), to describe the consistencies within inliers and inconsistencies within outliers. CF consists of top-ranked compatibility scores of a candidate to other correspondences, which purely relies on robust and rotation-invariant geometric constraints. We then formulate the grouping problem as a classification problem for CF features, which is accomplished via a simple multilayer perceptron (MLP) network. Comparisons with nine state-of-the-art methods on four benchmarks demonstrate that: 1) CF is distinctive, robust, and rotation-invariant; 2) our CF-based method achieves the best overall performance and holds good generalization ability.
AB - We present a simple yet effective method for 3D correspondence grouping. The objective is to accurately classify initial correspondences obtained by matching local geometric descriptors into inliers and outliers. Although the spatial distribution of correspondences is irregular, inliers are expected to be geometrically compatible with each other. Based on such observation, we propose a novel feature representation for 3D correspondences, dubbed compatibility feature (CF), to describe the consistencies within inliers and inconsistencies within outliers. CF consists of top-ranked compatibility scores of a candidate to other correspondences, which purely relies on robust and rotation-invariant geometric constraints. We then formulate the grouping problem as a classification problem for CF features, which is accomplished via a simple multilayer perceptron (MLP) network. Comparisons with nine state-of-the-art methods on four benchmarks demonstrate that: 1) CF is distinctive, robust, and rotation-invariant; 2) our CF-based method achieves the best overall performance and holds good generalization ability.
KW - 3D correspondence grouping
KW - 3D point cloud
KW - Compatability feature
UR - http://www.scopus.com/inward/record.url?scp=85118217292&partnerID=8YFLogxK
U2 - 10.1007/978-3-030-88007-1_6
DO - 10.1007/978-3-030-88007-1_6
M3 - 会议稿件
AN - SCOPUS:85118217292
SN - 9783030880064
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 66
EP - 78
BT - Pattern Recognition and Computer Vision - 4th Chinese Conference, PRCV 2021, Proceedings
A2 - Ma, Huimin
A2 - Wang, Liang
A2 - Zhang, Changshui
A2 - Wu, Fei
A2 - Tan, Tieniu
A2 - Wang, Yaonan
A2 - Lai, Jianhuang
A2 - Zhao, Yao
PB - Springer Science and Business Media Deutschland GmbH
T2 - 4th Chinese Conference on Pattern Recognition and Computer Vision, PRCV 2021
Y2 - 29 October 2021 through 1 November 2021
ER -