Article
Computer Science, Artificial Intelligence
Jiayuan Li, Pengcheng Shi, Qingwu Hu, Yongjun Zhang
Summary: With the development of 3D matching technology, correspondence-based point cloud registration has become more popular. However, a drawback of 3D keypoint techniques is the high number of outliers they produce. This paper proposes a new method called quadratic-time GORE (QGORE) that is much more efficient than previous methods while maintaining robustness and optimality.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2023)
Article
Computer Science, Information Systems
Linlin Ge, Jieqing Feng
Summary: The proposed type-based outlier removal framework (TBORF) classifies outliers more elaborately by considering both the characteristics of the underlying point cloud and the outliers. This allows for the design of more targeted outlier removal methods, which have been shown to effectively remove various outliers and improve subsequent digital geometry processing operations.
INFORMATION SCIENCES
(2021)
Article
Computer Science, Artificial Intelligence
Jiayuan Li
Summary: In this paper, a polynomial time outlier removal method is proposed, aiming to reduce the input set into a smaller one with a lower outlier rate based on bound principle. A scale-adaptive Cauchy estimator (SA-Cauchy) is introduced for further optimization. Extensive experiments demonstrate that the proposed method is robust at outlier rates above 99 percent and significantly faster than its competitors.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2022)
Article
Computer Science, Artificial Intelligence
Fan Yang, Zhi Chen, Kun Sun, Liman Liu, Wenbing Tao
Summary: This study proposes a robust consensus-aware network for removing outlier correspondences in feature-based point cloud registration. The method learns distinctive features for each correspondence by mining the global consensus and predicts the rigid transformation to align the point clouds. It achieves high registration accuracy and efficiency through feature similarity and multi-level context.
Article
Engineering, Multidisciplinary
Haonan Pei, Puyu Zhang, Sizhe Du, Ming Luo
Summary: This article proposes a point cloud preprocessing method for the inspection of metallic W-ring, which effectively removes noise and repairs missing data. Experimental results demonstrate the advantage of the proposed method.
Article
Geography, Physical
Maxime Kirgo, Guillaume Terrasse, Guillaume Thibault, Maks Ovsjanikov
Summary: In this paper, we propose a method to address the problem of reflection-induced outlier detection in laser acquisition of large-scale point clouds. We introduce a new dataset tailored for this task and demonstrate the utility of deep learning based semantic segmentation architectures for capturing non-local dependencies. We also show that incorporating additional non-local cues such as laser intensity and computed visibility signal significantly improves performance. Our ReVISOR pipeline outperforms existing baselines on real-world data.
ISPRS JOURNAL OF PHOTOGRAMMETRY AND REMOTE SENSING
(2023)
Article
Geography, Physical
Xingyu Jiang, Yang Wang, Aoxiang Fan, Jiayi Ma
Summary: In this paper, the authors propose a deep technique, called GANet, for recovering camera pose from two-view images. Unlike existing methods, GANet uses a graph attention operation to capture contextual/geometrical relations among corrupted correspondences, thus better extracting underlying geometry information. Moreover, the authors also propose a lightweight implementation for graph attention, named Sparse GANet, to reduce memory and computational requests. Experimental results demonstrate the superiority of GANet over other methods on various challenging datasets.
ISPRS JOURNAL OF PHOTOGRAMMETRY AND REMOTE SENSING
(2022)
Article
Computer Science, Artificial Intelligence
Ying Li, Huankun Sheng
Summary: As a representation for objects, 3D point cloud has gained attention due to its simplicity, flexibility, and effectiveness. However, raw point clouds often contain noise and outliers, which hinder downstream tasks. Point cloud cleaning is crucial in geometry processing workflows, but existing techniques require multiple models. In this paper, a novel single-stage point cloud cleaning network, SSPCN, is proposed to simultaneously remove outliers and denoise the point cloud. SSPCN outperforms state-of-the-art techniques in terms of quantitative metrics and visual quality.
PATTERN RECOGNITION
(2023)
Article
Chemistry, Multidisciplinary
Shijie Su, Chao Wang, Ke Chen, Jian Zhang, Hui Yang
Summary: Through the use of a novel MPCR-Net for multiple partial point cloud registration networks, this study demonstrates an improved accuracy and robustness in 3D point cloud registration by utilizing deep learning and rigid body transformation matrices.
APPLIED SCIENCES-BASEL
(2021)
Article
Construction & Building Technology
Feng Li, Wenzhong Shi, Yunlin Tu, Hua Zhang
Summary: An automatic framework with three steps is proposed to preprocess indoor point cloud, achieving successful reorientation and preservation of indoor points.
JOURNAL OF BUILDING ENGINEERING
(2023)
Article
Computer Science, Artificial Intelligence
Li Yan, Pengcheng Wei, Hong Xie, Jicheng Dai, Hao Wu, Ming Huang
Summary: This article introduces the high outlier ratio problem in correspondence-based point cloud registration with point feature techniques, and the issues of low efficiency, accuracy, and recall rate in current outlier removal methods. The authors propose an intuitive method to describe the 6-DOF curtailment process and an outlier removal strategy based on the reliability of the correspondence graph. Extensive experiments demonstrate that this method can effectively perform point cloud registration even with over 99% correspondence outlier ratio, and has better efficiency compared to existing methods.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2023)
Article
Geography, Physical
Xufei Wang, Zexin Yang, Xiaojun Cheng, Jantien Stoter, Wenbing Xu, Zhenlun Wu, Liangliang Nan
Summary: In this research, an automatic, robust, and efficient method for registering forest point clouds is proposed. The approach locates tree stems and matches them based on their relative spatial relationship to determine the registration transformation. The algorithm requires no extra tree attributes and can align point clouds of large forest environments. Additionally, a new benchmark dataset is introduced for the development and evaluation of forest point cloud registration methods.
ISPRS JOURNAL OF PHOTOGRAMMETRY AND REMOTE SENSING
(2023)
Article
Computer Science, Artificial Intelligence
Yujie Wang, Chenggang Yan, Yutong Feng, Shaoyi Du, Qionghai Dai, Yue Gao
Summary: Partial point cloud registration transforms partial scans into a common coordinate system, which is crucial for generating complete 3D shapes. Traditional registration methods struggle with small point cloud overlaps, but the STORM method utilizes structure information to accurately detect overlap and generate precise partial correspondences, achieving superior performance.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2023)
Article
Environmental Sciences
Shiming Li, Xuming Ge, Shengfu Li, Bo Xu, Zhendong Wang
Summary: In this paper, a systematic incremental registration method is proposed for successfully registering point cloud data from different sources. The robustness of this method is attributed to the elimination of noise in extracted linear features and its 2D incremental registration strategy. Experimental results demonstrate that this method can efficiently achieve data registration and has good generality.
Article
Computer Science, Software Engineering
Chen Wang, Yuhua Xu, Lin Wang, Chunming Li
Summary: This paper proposes a fast and effective algorithm for global registration of indoor colored point clouds based on the Manhattan-world assumption and color information. By limiting the number of rotation solutions and utilizing color information to reduce ambiguity induced by structural models, the algorithm is able to quickly complete registration tasks with comparable accuracy to other methods.
Article
Computer Science, Artificial Intelligence
Alvaro Parra Bustos, Tat-Jun Chin, Anders Eriksson, Hongdong Li, David Suter
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2016)
Article
Engineering, Environmental
Alvaro Parra, Julian M. Ortiz
STOCHASTIC ENVIRONMENTAL RESEARCH AND RISK ASSESSMENT
(2011)
Article
Geography, Physical
Zhipeng Cai, Tat-Jun Chin, Alvaro Parra Bustos, Konrad Schindler
ISPRS JOURNAL OF PHOTOGRAMMETRY AND REMOTE SENSING
(2019)
Proceedings Paper
Computer Science, Artificial Intelligence
Alvaro Parra Bustos, Tat-Jun Chin
2015 IEEE INTERNATIONAL CONFERENCE ON COMPUTER VISION (ICCV)
(2015)
Proceedings Paper
Computer Science, Artificial Intelligence
Alvaro Parra Bustos, Tat-Jun Chin, David Suter
2014 IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR)
(2014)