首页 > 论文 > 激光与光电子学进展 > 57卷 > 20期(pp:201102--1)

基于卷积神经网络结合改进Harris-SIFT的点云配准方法

Point Cloud Registration Method Based on Combination of Convolutional Neural Network and Improved Harris-SIFT

  • 摘要
  • 论文信息
  • 参考文献
  • 被引情况
  • PDF全文
分享:

摘要

针对传统点云配准方法在处理大型点云模型时存在计算量大、效率低和移动扫描配准实时性较差等问题,提出基于卷积神经网络结合改进Harris-SIFT(Scale Invariant Feature Transform)的点云配准方法。首先改进Harris-SIFT算法,使其可以提取三维空间中点云模型的稳定关键点。进而将关键点的加权邻接矩阵作为卷积神经网络的输入特征图,实现源点云和目标点云关键点的预测匹配。然后基于匹配的关键点,采用迭代最近点(ICP)算法实现点云数据的精配准。相较于传统的点对点配准,所提方法不需要生成对应关系的点描述符,解决全局搜索开销大的问题。实验结果表明,相较于ICP算法,所提方法能够较好地完成即时点云配准,且计算量小,耗时短,效率高。

Abstract

Considering the large amount of calculation, low efficiency, and poor real-time performance of mobile scanning registration when using the traditional point cloud registration method to process large point cloud models, a point cloud registration method based on the convolution neural network combined with the improved Harris-SIFT (Scale Invariant Feature Transform) is proposed. First, the Harris-SIFT algorithm is improved so that it can extract the stable key points of a point cloud model in three-dimensional space. Second, the weighted adjacency matrix of the key points is used as the input feature map for the convolutional neural network. This allows for prediction matching of the key points from the source point cloud and the target point cloud. Then, based on the key points of the matching, the iterative closest point (ICP) algorithm is used to achieve precise registration of the point cloud data. Compared with the traditional point-to-point registration, the proposed method does not need to generate corresponding point descriptors, thus avoiding the problem of a high global search overhead. The experimental results reveal that when compared with the ICP algorithm, the proposed method can better complete real-time point cloud registration, needs minimal calculation, takes a short amount of time, and is highly efficient.

广告组1 - 空间光调制器+DMD
补充资料

中图分类号:TP391.9

DOI:10.3788/LOP57.201102

所属栏目:成像系统

基金项目:国家自然科学基金、西安建筑科技大学基础研究基金;

收稿日期:2019-12-27

修改稿日期:2020-02-25

网络出版日期:2020-10-01

作者单位    点击查看

李昌华:西安建筑科技大学信息与控制工程学院, 陕西 西安 710055
史浩:西安建筑科技大学信息与控制工程学院, 陕西 西安 710055
李智杰:西安建筑科技大学信息与控制工程学院, 陕西 西安 710055

联系人作者:李智杰(lizhijie@xauat.edu.cn)

备注:国家自然科学基金、西安建筑科技大学基础研究基金;

【1】Tu D W, Lan H, Zhang X. Digital three-dimensional reconstruction technology of cultural relics [J]. Laser & Optoelectronics Progress. 2019, 56(19): 191504.
屠大维, 兰浩, 张曦. 文物数字化三维重建技术研究 [J]. 激光与光电子学进展. 2019, 56(19): 191504.

【2】Jiang J, Cheng J, Chen X L. Registration for 3-D point cloud using angular-invariant feature [J]. Neurocomputing. 2009, 72(16/17/18): 3839-3844.

【3】Yang J L, Li H D, Jia Y D. Go-ICP: solving 3D registration efficiently and globally optimally . [C]∥2013 IEEE International Conference on Computer Vision, December 1-8, 2013, Sydney, NSW, Australia. New York: IEEE. 2013, 1457-1464.

【4】Drost B, Ulrich M, Navab N, et al. Model globally, match locally: efficient and robust 3D object recognition . [C]∥2010 IEEE Computer Society Conference on Computer Vision and Pattern Recognition, June 13-18, 2010, San Francisco, CA, USA. New York: IEEE. 2010, 998-1005.

【5】Liang J G, Chen M L, Ma H. Registration of terrestrial laser scanning data based on projection distribution entropy [J]. Laser & Optoelectronics Progress. 2019, 56(13): 131501.
梁建国, 陈茂霖, 马红. 基于投影分布熵的地面激光点云自动配准方法 [J]. 激光与光电子学进展. 2019, 56(13): 131501.

【6】Besl P J. McKay N D. Method for registration of 3-D shapes [J]. Proceedings of SPIE. 1992, 1611: 586-606.

【7】Wang Y J, Lian T F, Wu M M, et al. Point cloud registration based on octree and KD-tree index [J]. Engineering of Surveying and Mapping. 2017, 26(8): 35-40.
王育坚, 廉腾飞, 吴明明, 等. 基于八叉树与KD树索引的点云配准方法 [J]. 测绘工程. 2017, 26(8): 35-40.

【8】Wang C, Shu Q, Yang Y X, et al. Quick registration algorithm of point clouds using structure feature [J]. Acta Optica Sinica. 2018, 38(9): 0911005.
王畅, 舒勤, 杨赟秀, 等. 利用结构特征的点云快速配准算法 [J]. 光学学报. 2018, 38(9): 0911005.

【9】Huang Y, Da F P, Tao H J. An automatic registration algorithm for point cloud based on feature extraction [J]. Chinese Journal of Lasers. 2015, 42(3): 0308002.
黄源, 达飞鹏, 陶海跻. 一种基于特征提取的点云自动配准算法 [J]. 中国激光. 2015, 42(3): 0308002.

【10】Tang Z R, Liu M Z, Jiang Y, et al. Point cloud registration algorithm based on canonical correlation analysis [J]. Chinese Journal of Lasers. 2019, 46(4): 0404006.
唐志荣, 刘明哲, 蒋悦, 等. 基于典型相关分析的点云配准算法 [J]. 中国激光. 2019, 46(4): 0404006.

【11】Wang Z J. A CNN regression approach for real-time 2D/3D registration [J]. IEEE Transactions on Medical Imaging. 2016, 35(5): 1352-1363.

【12】Díez Y, Roure F, Lladó X, et al. A qualitative review on 3D coarse registration methods [J]. ACM Computing Surveys. 2015, 47(3): 45.Díez Y, Roure F, Lladó X, et al. A qualitative review on 3D coarse registration methods [J]. ACM Computing Surveys. 2015, 47(3): 45.

【13】Tao H J, Da F P. Automatic registration algorithm for the point clouds based on the normal vector [J]. Chinese Journal of Lasers. 2013, 40(8): 0809001.
陶海跻, 达飞鹏. 一种基于法向量的点云自动配准方法 [J]. 中国激光. 2013, 40(8): 0809001.

【14】Lowe D G. Distinctive image features from scale-invariant keypoints [J]. International Journal of Computer Vision. 2004, 60(2): 91-110.

【15】Shu C X, He Y T, Sun Q K. Point cloud registration based on convolutional neural network [J]. Laser & Optoelectronics Progress. 2017, 54(3): 031001.
舒程珣, 何云涛, 孙庆科. 基于卷积神经网络的点云配准方法 [J]. 激光与光电子学进展. 2017, 54(3): 031001.

【16】Wang G P, Duan M, Niu C Y, et al. Stochastic gradient descent algorithm based on convolution neural network [J]. Computer Engineering and Design. 2018, 39(2): 441-445, 462.
王功鹏, 段萌, 牛常勇. 基于卷积神经网络的随机梯度下降算法 [J]. 计算机工程与设计. 2018, 39(2): 441-445, 462.

【17】Li C L, Dian S Y, Liu H L. Point cloud registration based on improved dynamic differential evolution algorithm [J]. Electronics Optics & Control. 2019, 26(3): 59-64.
李传龙, 佃松宜, 刘海亮. 基于改进动态差分进化算法的点云配准 [J]. 电光与控制. 2019, 26(3): 59-64.

引用该论文

Li Changhua,Shi Hao,Li Zhijie. Point Cloud Registration Method Based on Combination of Convolutional Neural Network and Improved Harris-SIFT[J]. Laser & Optoelectronics Progress, 2020, 57(20): 201102

李昌华,史浩,李智杰. 基于卷积神经网络结合改进Harris-SIFT的点云配准方法[J]. 激光与光电子学进展, 2020, 57(20): 201102

您的浏览器不支持PDF插件,请使用最新的(Chrome/Fire Fox等)浏览器.或者您还可以点击此处下载该论文PDF