基于主成分分析和特征图匹配的点云配准方法

doi:10.11959/j.issn.2096-6652.202340

智能科学与技术学报 ›› 2023, Vol. 5 ›› Issue (4): 543-552.doi: 10.11959/j.issn.2096-6652.202340

• 学术论文 • 上一篇

基于主成分分析和特征图匹配的点云配准方法

郑伟斌¹, 练国富¹(), 张学明¹, 郭方²

^1.福建理工大学机械与汽车工程学院，福建福州 350118
^2.福建理工大学计算机科学与数学学院，福建福州 350118

收稿日期:2022-12-07 修回日期:2023-06-02 出版日期:2023-12-15 发布日期:2023-12-15
通讯作者: 练国富 E-mail:gflian@mail.ustc.edu.cn
作者简介:郑伟斌（1998- ），男，福建理工大学机械与汽车工程学院硕士生，主要研究方向为计算机视觉与三维重建。
练国富（1980- ），男，博士，福建理工大学机械与汽车工程学院教授，主要研究方向为增材制造技术与机器视觉。
张学明（1996- ），男，福建理工大学机械与汽车工程学院硕士生，主要研究方向为计算机视觉与深度学习。
郭方（1977- ），男，博士，福建理工大学计算机科学与数学学院副教授，主要研究方向为计算机视觉、大数据处理与分析。
基金资助:
福建省科技重大专项专题项目(2020HZ03018)

Point cloud registration method based on principal component analysis and feature map matching

Weibin ZHENG¹, Guofu LIAN¹(), Xueming ZHANG¹, Fang GUO²

^1.School of Mechanical and Automotive Engineering, Fujian University of Technology, Fuzhou 350118, China
^2.School of Computer Science and Mathematics, Fujian University of Technology, Fuzhou 350118, China

Received:2022-12-07 Revised:2023-06-02 Online:2023-12-15 Published:2023-12-15
Contact: Guofu LIAN E-mail:gflian@mail.ustc.edu.cn
Supported by:
Science and Technology Major Project of Fujian Province(2020HZ03018)

摘要/Abstract

摘要：

由于点云模型存在不同程度的重叠，点云配准容易出现特征匹配错误、配准难度大等问题。因此，提出了一种基于主成分分析和特征图匹配的点云配准方法。配准前，首先采用带主轴校正的主成分分析方法进行点云初始位姿调整，建立KD树进行重叠区域搜索。其次，根据两幅点云的重叠区域计算采样点的快速点特征直方图特征，进行点云特征图匹配以及裁剪迭代最近点精配准。在现有数据集以及实际扫描模型上进行配准实验，实验结果表明该方法的稳定性好，精度更高，相较于其他算法精度能提高25%以上。

关键词: 重叠区域, KD树, 图匹配, 裁剪迭代最近点, 点云配准

Abstract:

Due to varying degrees of overlap in point cloud models, point cloud registration is prone to problems, such as feature matching errors and high difficulty in registration. Therefore, a point cloud registration method based on principal component analysis and feature map matching is proposed. Before registration, the principal component analysis method with spindle correction was used to adjust the initial pose, then the K-dimensional tree was established to search the overlapping area. Secondly, the fast point feature histograms features of the sampling points were calculated according to the overlapping area of the two-point cloud, and the point cloud feature graph matching and trimmed iterative closest point (TrICP) fine registration were performed. Registration experiments were carried out according to the existing datasets and the actual scanning model. The experimental results show that the method has good stability and higher accuracy, and the accuracy can be improved by more than 25% compared with other algorithms.

Key words: overlapping region, K-dimension tree, graph matching, TrICP, point cloud registration

中图分类号:

TP391

郑伟斌, 练国富, 张学明, 等. 基于主成分分析和特征图匹配的点云配准方法[J]. 智能科学与技术学报, 2023, 5(4): 543-552.

Weibin ZHENG, Guofu LIAN, Xueming ZHANG, et al. Point cloud registration method based on principal component analysis and feature map matching[J]. Chinese Journal of Intelligent Science and Technology, 2023, 5(4): 543-552.

图/表 18

图1

图2

图3

图4

图5

图6

表1

图7

图8

图9

表2

点云数据配准误差分析"

点云数据	配准算法	配准误差MSE	时间/s
Bunny	ICP	$11.104 × 10 - 4$	32.1
	文献[8]	$1.922 × 10 - 4$	25.4
	文献[11]	$2.129 × 10 - 4$	19.2
	文献[16]	$2.562 × 10 - 4$	17.9
	FGM+TrICP	$1.816 × 10 - 4$	15.2
Dragon	ICP	$5.165 × 10 - 5$	28.7
	文献[8]	$6.080 × 10 - 5$	22.1
	文献[11]	$4.557 × 10 - 5$	18.9
	文献[16]	$6.921 × 10 - 5$	13.5
	FGM+TrICP	$3.964 × 10 - 5$	14.3
7-Scenes室内场景	ICP	$2.413 × 10 - 3$	42.7
	文献[8]	$3.458 × 10 - 3$	32.7
	文献[11]	$2.556 × 10 - 3$	25.1
	文献[16]	$2.388 × 10 - 3$	22.0
	FGM+TrICP	$2.056 × 10 - 3$	19.8

表2

图10

图11

表3

图12

图13

图14

表4

齿轮箱体点云配准误差分析"

点云数据	配准算法	配准误差MSE	时间/s
齿轮箱体	ICP	$7.411 × 10 - 4$	39.8
	文献[8]	$4.513 × 10 - 4$	32.7
	文献[11]	$5.253 × 10 - 4$	21.4
	文献[16]	$5.770 × 10 - 4$	16.5
	FGM+TrICP	$2.593 × 10 - 4$	17.2

表4

参考文献 23

1	李博洋, 刘思健, 崔明月, 等. 基于最小回环检测的多车协同SLAM框架[J]. 电子学报, 2021, 49(11): 2241-2250.
	LI B Y, LIU S J, CUI M Y, et al. Multi-vehicle collaborative SLAM framework for minimum loop detection[J]. Acta Electronica Sinica, 2021, 49(11): 2241-2250.
2	DONG Z, LIANG F X, YANG B S, et al. Registration of large-scale terrestrial laser scanner point clouds: a review and benchmark[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2020, 163: 327-342.
3	BESL P J, MCKAY N D. Method for registration of 3-D shapes[C]//Proceedings of SPIE 1611, Sensor Fusion IV: Control Paradigms and Data Structures. Washington: SPIE, 1992, 1611: 586-606.
4	ZHONG L S, YING J, YANG H M, et al. Triple screening point cloud registration method based on image and geometric features[J]. Optik, 2021, 246: 167763.
5	张旭春, 周洪军, 郑津津, 等. 基于多尺度特征及点距离约束的点云配准[J]. 激光与光电子学进展, 2021, 58(24): 325-333.
	ZHANG X C, ZHOU H J, ZHENG J J, et al. Point cloud registration based on multi-scale features and point distance constraints[J]. Laser & Optoelectronics Progress, 2021, 58(24): 325-333.
6	XU G X, PANG Y J, BAI Z X, et al. A fast point clouds registration algorithm for laser scanners[J]. Applied Sciences, 2021, 11(8): 3426.
7	ZAGANIDIS A, MAGNUSSON M, DUCKETT T, et al. Semantic-assisted 3D normal distributions transform for scan registration in environments with limited structure[C]//Proceedings of 2017 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). Piscataway: IEEE Press, 2017: 4064-4069.
8	荆路, 武斌, 李先帅. 基于SAC-IA和NDT融合的点云配准方法[J]. 大地测量与地球动力学, 2021, 41(4): 378-381.
	JING L, WU B, LI X S. Point cloud registration method based on SAC-IA and NDT fusion[J]. Journal of Geodesy and Geodynamics, 2021, 41(4): 378-381.
9	杜辉, 郑长亮, 苗春雨, 等. 一种动态特征匹配的部分重叠点云配准方法[J]. 电信科学, 2021, 37(4): 97-107.
	DU H, ZHENG C L, MIAO C Y, et al. A partial overlapping point cloud registration method based on dynamic feature matching[J]. Telecommunications Science, 2021, 37(4): 97-107.
10	李鑫, 莫思特, 黄华, 等. 自动计算重叠度的多源点云配准方法[J]. 红外与激光工程, 2021, 50(12): 260-268.
	LI X, MO S T, HUANG H, et al. Multi-source point cloud registration method for automatically calculating overlap degree[J]. Infrared and Laser Engineering, 2021, 50(12): 260-268.
11	LI J, QIAN F, CHEN X F. Point cloud registration algorithm based on overlapping region extraction[J]. Journal of Physics: Conference Series, 2020, 1634(1): 012012.
12	张元, 李晓燕, 韩燮. 一种低重叠率的三维点云配准方法[J]. 激光与光电子学进展, 2021, 58(8):162-171.
	ZHANG Y, LI X, HAN X. Three-dimensional point cloud registration method with low overlap rate[J]. Laser & Optoelectronics Progress, 2021, 58(8):162-171.
13	POIESI F, BOSCAINI D. Learning general and distinctive 3D local deep descriptors for point cloud registration[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2023, 45(3): 3979-3985.
14	WANG Y, SOLOMON J. Deep closest point: learning representations for point cloud registration[C]//Proceedings of 2019 IEEE/CVF International Conference on Computer Vision (ICCV). Piscataway: IEEE Press, 2020: 3522-3531.
15	WU Z R, SONG S R, KHOSLA A, et al. 3D ShapeNets: a deep representation for volumetric shapes[C]//Proceedings of 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Piscataway: IEEE Press, 2015: 1912-1920.
16	FU K X, LIU S L, LUO X Y, et al. Robust point cloud registration framework based on deep graph matching[C]//Proceedings of 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Piscataway: IEEE Press, 2021: 8889-8898.
17	HUANG S Y, GOJCIC Z, USVYATSOV M, et al. PREDATOR: registration of 3D point clouds with low overlap[C]//Proceedings of 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Piscataway: IEEE Press, 2021: 4265-4274.
18	CHENG D Y, ZHAO D J, ZHANG J C, et al. PCA-based denoising algorithm for outdoor lidar point cloud data[J]. Sensors, 2021, 21(11): 3703.
19	王来强, 包静. 附加主方向判定的PCA点云数据初始配准算法[J]. 测绘地理信息, 2021, 46(6):59-62, 66.
	WANG L Q, BAO J. Initial registration algorithm for PCA point cloud data with additional main direction decision[J]. Journal of Geomatics, 2021, 46(6): 59-62, 66.
20	LIU J, ZHU J W, YANG J L, et al. Three-dimensional point cloud registration based on ICP algorithm employing K-D tree optimization[C]//Proceedings of Eighth International Conference on Digital Image Processing (ICDIP 2016). Washington: SPIE, 2016, 10033: 872-876.
21	RUSU R B, BLODOW N, BEETZ M. Fast point feature histograms (FPFH) for 3D registration[C]//Proceedings of 2009 IEEE International Conference on Robotics and Automation. Piscataway: IEEE Press, 2009: 3212-3217.
22	CHETVERIKOV D, SVIRKO D, STEPANOV D, et al. The trimmed iterative closest point algorithm[C]//Proceedings of 2002 International Conference on Pattern Recognition. Piscataway: IEEE Press, 2002: 545-548.
23	LI J Y, HU Q W, AI M Y. Point cloud registration based on one-point RANSAC and scale-annealing biweight estimation[J]. IEEE Transactions on Geoscience and Remote Sensing, 2021, 59(11): 9716-9729.

模型	重叠率	点数/个	质心位置
Bunny	31%	Q: 40 097	0.0104，0.0984，0.0606
Bunny	31%	P: 30 379	-0.0064，0.1027，0.0064
Dragon	44%	Q: 41 841	-0.0044，0.1157，0.0058
Dragon	44%	P: 22 092	0.0045，0.1145，0.0114
7-Scenes 室内场景	58%	Q: 291 605	0.1772，-0.3775，2.0563
7-Scenes 室内场景	58%	P: 268 376	0.5535，-0.1114，2.0154

点云数据	配准算法	配准误差MSE	时间/s
Bremen city	ICP	7.208	185.7
	文献[8]	1.055	115.6
	文献[11]	0.566	75.3
	文献[16]	1.489	68.6
	FGM+TrICP	0.102	70.2
Wue city	ICP	12.054	192.3
	文献[8]	1.244	105.7
	文献[11]	0.987	77.9
	文献[16]	1.442	69.1
	FGM+TrICP	0.635	67.5

[1]	崔文成, 王可丽, 邵虹. 基于稠密块和注意力机制的肺部病理图像异常细胞分割[J]. 智能科学与技术学报, 2023, 5(4): 525-534.
[2]	唐炜, 谭啸, 孙宇, 严家鹏, 严光锐. 基于A*与动态窗口法的物料传输平台路径规划研究[J]. 智能科学与技术学报, 2023, 5(4): 515-524.
[3]	方祯祺, 李雪, 莫红. 基于改进EfficientNet的乳腺肿瘤诊断[J]. 智能科学与技术学报, 2023, 5(4): 505-514.
[4]	靳聪, 王洁, 郭子淳, 王晶. 融合音画同步的唇形合成研究[J]. 智能科学与技术学报, 2023, 5(3): 397-405.
[5]	项凤涛, 苏炯铭, 谷学强, 张万鹏. 基于智能体建模的新冠肺炎疫情传播问题研究[J]. 智能科学与技术学报, 2023, 5(1): 51-57.
[6]	顾宙瑜, 於跃成, 者甜甜. Rapider-YOLOX：高效的轻量级目标检测网络[J]. 智能科学与技术学报, 2023, 5(1): 92-103.
[7]	丛晓峰, 桂杰, 章军. 基于视觉Transformer的多损失融合水下图像增强网络[J]. 智能科学与技术学报, 2022, 4(4): 522-532.
[8]	于航, 付彦伟, 姜柏言, 薛向阳. 基于少量图像的三维重建综述[J]. 智能科学与技术学报, 2022, 4(4): 544-559.
[9]	陈妍, 罗雪琴, 梁伟, 谢永芳. 基于情感信息融合注意力机制的抑郁症识别[J]. 智能科学与技术学报, 2022, 4(4): 600-609.
[10]	张永强, 宋美霖, 刘天虎, 满梦华. 混合骨骼特征的三帧间差分手势识别方法研究[J]. 智能科学与技术学报, 2022, 4(4): 592-599.
[11]	郭超, 鲁越, 王晓, 易达, 王虓, 王飞跃. 人机物CPSS智能融合的平行创作架构与关键技术研究[J]. 智能科学与技术学报, 2022, 4(3): 344-354.
[12]	郁洲, 毕敬, 苑海涛. 基于改进DQN算法的复杂海战场路径规划方法[J]. 智能科学与技术学报, 2022, 4(3): 418-425.
[13]	栗仁武, 张凌霄, 高林, 李淳芃, 蒋浩. 基于点云的类级别物体姿态估计[J]. 智能科学与技术学报, 2022, 4(2): 246-254.
[14]	陈名杨, 张文, 陈湘楠, 周虹廷, 陈华钧. 群体知识图谱：分布式知识迁移与联邦式图谱推理[J]. 智能科学与技术学报, 2022, 4(1): 55-64.
[15]	余跃, 刘欣, 蒋芳清, 张晗, 王晖, 曾炜. 支持鹏程系列开源大模型应用生态演化的可持续学习能力探索[J]. 智能科学与技术学报, 2022, 4(1): 97-108.

基于主成分分析和特征图匹配的点云配准方法

Point cloud registration method based on principal component analysis and feature map matching

在线阅读

PDF下载

可视化

摘要/Abstract

引用本文

使用本文

图/表 18

参考文献 23

相关文章 15

Metrics

推荐阅读 0