面向盾牌形貌重建的双层采样点云粗配准方法

doi:10.16180/j.cnki.issn1007-7820.2024.07.008

电子科技 ›› 2024, Vol. 37 ›› Issue (7): 60-65.doi: 10.16180/j.cnki.issn1007-7820.2024.07.008

面向盾牌形貌重建的双层采样点云粗配准方法

万晟睿¹, 周志峰¹^,², 吴明晖¹^,², 王立端³, 周围⁴

1.上海工程技术大学机械与汽车工程学院,上海 201620
2.上海市大型构件智能制造机器人技术协同创新中心,上海 201620
3.上海司南卫星导航技术股份有限公司,上海 201801
4.陆军装备部驻湘潭地区军事代表室,湖南湘潭 411100

收稿日期:2023-02-05 出版日期:2024-07-15 发布日期:2024-07-17
作者简介:万晟睿(1997-),男,硕士研究生。研究方向:单目结构光三维重建。
周志峰(1976-),男,博士,副教授。研究方向:计算机测控、自动导航驾驶、机器视觉与运动控制、嵌入式与信号处理。
吴明晖(1973-),男,博士,讲师。研究方向:机器人环境感知与控制、特种机器人机构创新。
基金资助:
上海市优秀学术/技术带头人计划(22XD1433500)

Coarse Registration Method of Double-Layer Sampling Point Cloud for Shield Topography Reconstruction

WAN Shengrui¹, ZHOU Zhifeng¹^,², WU Minghui¹^,², WANG Liduan³, ZHOU Wei⁴

1. School of Mechanical and Automotive Engineering, Shanghai University of Engineering Science, Shanghai 201620, China
2. Shanghai Collaborative Innovation Center of Intelligent Manufacturing Robot Technology for Large Components, Shanghai 201620, China
3. ComNav Technology Co.,Ltd., Shanghai 201801, China
4. Military Representative Office of the Army Equipment Department in Xiangtan District,Xiangtan 411100, China

Received:2023-02-05 Online:2024-07-15 Published:2024-07-17
Supported by:
Shanghai Excellent Academic/Technical Leader Program(22XD1433500)

摘要/Abstract

摘要：

针对盾牌表面形貌三维重建过程中点云粗配准精度低和耗时长的问题,文中基于传统RANSAC(Random Sample Consensus)点云粗配准算法框架提出了一种双层采样点云粗配准算法。在第1层每次迭代中,采用单点采样进行刚性约束来减小对应集的大小,将此对应集作为第2层的“内点候选集”。算法第2层执行两个点的连续随机采样以及计算各自的最小模型,通过迭代最大化一致性集合并利用最小二乘法求解出最佳刚体变换矩阵。在实验阶段,采用6组降采样程度不同的盾牌点云,使用RANSAC算法和本文算法进行点云粗配准对比实验。实验结果表明,所提算法在粗配准速度和粗配准精度两方面都优于RANSAC算法,配准速度约为RANSAC算法的两倍,均方根误差为10^-3 mm,比RANSAC算法的精度提高近400倍。

关键词: 形貌重建, 粗配准, RANSAC, 双层采样, 单点采样, 刚性约束, 一致性集合, 盾牌

Abstract:

In view of the problems of low precision and time-consuming point cloud coarse registration in the process of 3D reconstruction of shield surface topography, this study proposes a double-layer sampling point cloud coarse registration algorithm based on the traditional RANSAC(Random Sample Consensus) point cloud coarse registration algorithm framework. In each iteration of the first layer of the proposed algorithm, the single-point sampling is used for rigid constraints to reduce the size of the corresponding set, which is regarded as the "interior point candidate set" of the second layer. The second layer of the algorithm performs continuous random sampling of two points and computes their respective minimum models. The optimal rigid body transformation matrix is obtained by iteratively maximizing the consistent set and using the least square method. In the experimental stage, 6 groups of shield point clouds with different degrees of down sampling are used, and the RANSAC algorithm and the proposed algorithm are used to conduct a point cloud coarse registration comparison experiment. Experimental results show that the proposed algorithm is superior to the RANSAC algorithm in both rough registration speed and rough registration accuracy. The registration speed is about twice that of the RANSAC algorithm, and the root-mean-square error of the proposed algorithm is 10^-3 mm, which is nearly 400 times higher than the RANSAC algorithm.

Key words: topography reconstruction, rough registration, RANSAC, double-layer sampling, single point sampling, rigid constraint, consensus set, shield

中图分类号:

TP391

万晟睿, 周志峰, 吴明晖, 王立端, 周围. 面向盾牌形貌重建的双层采样点云粗配准方法[J]. 电子科技, 2024, 37(7): 60-65.

WAN Shengrui, ZHOU Zhifeng, WU Minghui, WANG Liduan, ZHOU Wei. Coarse Registration Method of Double-Layer Sampling Point Cloud for Shield Topography Reconstruction[J]. Electronic Science and Technology, 2024, 37(7): 60-65.

图/表 11

图1

图2

图3

图4

图5

图6

图7

图8

图9

表1

图10

参考文献 18

[1]	曾祥磊. 基于几何特征的三维点云配准算法研究[D]. 济南: 山东大学, 2020:1-8.
	Zeng Xianglei. Research on 3D point cloud registration algorithm based on geometric features[D]. Jinan: Shandong University, 2020:1-8.
[2]	王庆闪, 张军, 刘元盛, 等. 基于NDT与ICP结合的点云配准算法[J]. 计算机工程与应用, 2020, 56(7):88-95. doi: 10.3778/j.issn.1002-8331.1904-0174
	Wang Qingshan, Zhang Jun, Liu Yuansheng, et al. Point cloud registration algorithm based on combination of NDT and ICP[J]. Computer Engineering and Applications, 2020, 56(7):88-95. doi: 10.3778/j.issn.1002-8331.1904-0174
[3]	Besl P J. A method for registration of 3D shapes[J]. IEEE Trans on PAMI, 1992, 14(2):239-256.
[4]	Rusu R B, Holzbach A, Blodow N, et al. Fast geometric point labeling using conditional random fields[C]. Saint Louis: IEEE/RSJ International Conference on Intelligent Robots and Systems, 2009:7-12.
[5]	Tangelder J, Veltkamp R C. Polyhedral model retrieval using weighted point sets[C]. Seoul: Shape Modeling International, 2003:119-129.
[6]	曲天伟, 安波, 陈桂兰. 改进的RANSAC算法在图像配准中的应用[J]. 计算机应用, 2010(7):4-6.
	Qu Tianwei, An Bo, Chen Guilan. Application of improved RANSAC algorithm to image registration[J]. Journal of Computer Applications, 2010(7):4-6.
[7]	辛明, 张苗辉. 基于SURF的红外成像末制导目标跟踪算法[J]. 光电子·激光, 2012, 23(8):1597-1603.
	Xin Ming, Zhang Miaohui. SURF based infrared imaging terminal guidance target tracking algorithm[J]. Journal of Optoelectronics·Laser, 2012, 23(8):1597-1603.
[8]	刘美菊, 王旭东, 李凌燕, 等. 改进的RANSAC算法在三维点云配准中的应用[J]. 激光与光电子学进展, 2018, 55(10):7-10.
	Liu Meiju, Wang Xudong, Li Lingyan, et al. Improved random sampling consistency algorithm employed in three-dimensional point cloud registration[J]. Laser and Optoelectronics Progress, 2018, 55(10):7-10.
[9]	Briales J, Gonzalez-Jimenez J. Convex global 3D registration with lagrangian duality[C]. Honolulu: IEEE Conference on Computer Vision and Pattern Recognition, 2017:5612-5621.
[10]	张志鹏, 邵学君, 庞庆. 基于三维激光扫描的铁路罐车点云优化处理方法[J]. 应用光学, 2022, 43(3):460-465. doi: 10.5768/JAO202243.0302006
	Zhang Zhipeng, Shao Xuejun, Pang Qing. Optimized processing method of point cloud of railway tank car based on 3D laser scanning[J]. Journal of Applied Optics, 2022, 43(3):460-465. doi: 10.5768/JAO202243.0302006
[11]	Drost B, Ulrich M, Navab N, et al. Model globally,match locally:efficient and robust 3D object recognition[C]. San Francisco: IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2010:998-1005.
[12]	Ji Z, Singh S. Visual-lidar odometry and mapping:Low-drift, robust,and fast[C]. Seattle: IEEE International Conference on Robotics and Automation, 2015:2174-2181.
[13]	Yu W, Hui Z. 3D reconstruction of indoor scenes based on feature and graph optimization[C]. Hangzhou: International Conference on Virtual Reality and Visualization, 2016:126-132.
[14]	Henry P, Krainin M, Herbst E, et al. RGB-D mapping:Using kinect-style depth cameras for dense 3D modeling of indoor environments[J]. International Journal of Robotics Research, 2012(5):31-38.
[15]	李思远, 刘瑾, 杨海马, 等. 分两阶段变换坐标的点云粗配准算法[J]. 激光与光电子学进展, 2022, 59(16):127-134.
	Li Siyuan, Liu Jin, Yang Haima, et al. Point cloud coarse registration algorithm basedon two-stage coordinate transformation[J]. Laser and Optoelectronics Progress, 2022, 59(16):127-134.
[16]	邓兵华, 张明. MATLAB环境下图像配准方法[J]. 电子科技, 2014, 27(1):139-142.
	Deng Binghua, Zhang Ming. Methods of image registration based on MATLAB[J]. Electronic Science and Technology, 2014, 27(1):139-142.
[17]	刘桥, 杨正坤, 李晗. 基于SURF算法的医学图像特征点匹配[J]. 电子科技, 2014, 27(5):145-148.
	Liu Qiao, Yang Zhengkun, Li Han. SURF algorithm for medical image based on feature points matching[J]. Electronic Science and Technology, 2014, 27(5):145-148.
[18]	陆尚鸿, 李文国. 基于点云边界质心的粗配准方法[J]. 电子科技, 2022, 35(4):53-59.
	Lu Shanghong, Li Wenguo. The point clould coarse registration method based on boundary centroid[J]. Electronic Science and Technology, 2022, 35(4):53-59.

配准算法	点云数	时间/s	均方根误差/mm
RANSAC算法	99 675	4.475	2.917
本文算法	99 675	2.175	0.328×10^-3
RANSAC算法	88 608	4.960	0.708
本文算法	88 608	2.170	2.339×10^-3
RANSAC算法	73 746	5.450	0.607
本文算法	73 746	2.091	2.331×10^-3
RANSAC算法	63 999	4.221	1.590
本文算法	63 999	1.623	2.798×10^-3
RANSAC算法	54 968	3.808	1.226
本文算法	54 968	1.345	1.724×10^-3
RANSAC算法	45 899	4.173	3.111
本文算法	45 899	1.204	0.898×10^-3

面向盾牌形貌重建的双层采样点云粗配准方法

Coarse Registration Method of Double-Layer Sampling Point Cloud for Shield Topography Reconstruction

RichHTML

PDF (PC)

赞

可视化

摘要/Abstract

引用本文

使用本文

图/表 11

参考文献 18

相关文章 2

Metrics

本文评价

推荐阅读 0

[1]	陆尚鸿,李文国. 基于点云边界质心的粗配准方法[J]. 电子科技, 2022, 35(4): 53-59.
[2]	王正军,姚一鸣,陈龙. 基于卷积神经网络的PCB板通用分拣算法[J]. 电子科技, 2020, 33(2): 60-65.