基于度量学习的回环检测描述子提升算法

doi:10.16451/j.cnki.issn1003-6059.202201005

Abstract
Figure/Table
References
Related Citation (9)

Download: PDF (1740 KB) HTML (1 KB)
Export: BibTeX | EndNote (RIS)

Abstract Loop detection is an important part of simultaneous localization and mapping (SLAM). In most of the loop detection algorithms, feature descriptors are extracted from data frames, and loops are searched through the Euclidean distance between the descriptors. However, feature enhancement is not conducted on the extracted feature descriptors. In this paper, an algorithm of feature descriptor enhancement for loop detection based on metric learning is proposed. A lightweight algorithm module is designed to transform the feature space of the descriptors to enhance the distinguishing ability of the descriptors and improve the loop detection performance effectively. Pose and descriptors are combined to establish a triple dataset and thus the problem of fuzzy labels is solved. An idea of expanding the dataset is proposed to solve the problem of significantly insufficient loop samples. Based on triplet loss, the proposed loss function is adapted to the loop detection scene, and it is utilized to train a neural network module for feature space transformation. Experiments on KITTI and NCLT datasets show that the generalization ability of the proposed algorithm is strong.

Key words： Simultaneous Localization and Mapping(SLAM) Loop Detection Feature Descriptor Feature Enhancement Metric Learning

Received: 25 May 2021

Corresponding Authors: SHEN Huiliang, Ph. D., professor. His research interests include image processing and computer vision.

About author:: HAN Bin, master student. His research interests include lidar SLAM and machine learning.
LUO Lun, Ph. D. candidate. His research interests include lidar SLAM and lidar point cloud processing.
LIU Xiongwei, master student. His research interests include image processing and machine learning.

	Service

	E-mail this article
	Add to my bookshelf
	Add to citation manager
	E-mail Alert
	RSS
	Articles by authors
	HAN Bin
	LUO Lun
	LIU Xiongwei
	SHEN Huiliang

Cite this article:

HAN Bin,LUO Lun,LIU Xiongwei等. Feature Descriptor Enhancement for Loop Detection Based on Metric Learning[J]. Pattern Recognition and Artificial Intelligence, 2022, 35(1): 51-61.

URL:

http://manu46.magtech.com.cn/Jweb_prai/EN/10.16451/j.cnki.issn1003-6059.202201005 OR http://manu46.magtech.com.cn/Jweb_prai/EN/Y2022/V35/I1/51

[1] DUBÉ R, DUGAS D, STUMM E, et al. SegMatch: Segment Based Place Recognition in 3D Point Clouds // Proc of the IEEE International Conference on Robotics and Automation. Washington, USA: IEEE, 2017: 5266-5272.
[2] UY M A, LEE G H.PointNetVLAD: Deep Point Cloud Based Retrieval for Large-Scale Place Recognition // Proc of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Washington, USA: IEEE, 2018: 4470-4479.
[3] LIU Z, ZHOU S B, SUO C Z, et al. LPD-Net: 3D Point Cloud Learning for Large-Scale Place Recognition and Environment Analysis // Proc of the IEEE/CVF International Conference on Computer Vision. Washington, USA: IEEE, 2019: 2831-2840.
[4] CHARLES R Q, SU H, MO K, et al. PointNet: Deep Learning on Point Sets for 3D Classification and Segmentation // Proc of the IEEE Conference on Computer Vision and Pattern Recognition. Washington, USA: IEEE, 2017: 77-85.
[5] ARANDJELOVIC R, GRONAT P, TORII A, et al. NetVLAD: CNN Architecture for Weakly Supervised Place Recognition. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2018, 40(6): 1437-1451.
[6] HE L, WANG X L, ZHANG H.M2DP: A Novel 3D Point Cloud Descriptor and Its Application in Loop Closure Detection // Proc of the IEEE/RSJ International Conference on Intelligent Robots and Systems. Washington, USA: IEEE, 2016: 231-237.
[7] TOMBARI F, SALTI S, STEFANO L D.A Combined Texture-Shape Descriptor for Enhanced 3D Feature Matching // Proc of the 18th IEEE International Conference on Image Processing. Washington, USA: IEEE, 2011: 809-812.
[8] RUSU R B, BRADSKI G, THIBAUX R, et al. Fast 3D Recognition and Pose Using the Viewpoint Feature Histogram // Proc of the IEEE/RSJ International Conference on Intelligent Robots and Systems. Washington, USA: IEEE, 2010: 2155-2162.
[9] WOHLKINGER W, VINCZE M.Ensemble of Shape Functions for 3D Object Classification // Proc of the IEEE International Confe-rence on Robotics and Biomimetics. Washington, USA: IEEE, 2011: 2987-2992.
[10] KIM G, KIM A.Scan Context: Egocentric Spatial Descriptor for Place Recognition within 3D Point Cloud Map // Proc of the IEEE/RSJ International Conference on Intelligent Robots and Systems. Washington, USA: IEEE, 2018: 4802-4809.
[11] WANG Y, SUN Z Z, XU C Z, et al. LiDAR Iris for Loop-Closure Detection // Proc of the IEEE/RSJ International Conference on Intelligent Robots and Systems. Washington, USA: IEEE, 2020: 5769-5775.
[12] SCHROFF F, KALENICHENK D, PHILBIN J.FaceNet: A Unified Embedding for Face Recognition and Clustering // Proc of the IEEE Conference on Computer Vision and Pattern Recognition. Washington, USA: IEEE, 2015: 815-823.
[13] SHRIVASTAVA A, GUPTA A, GIRSHICK R.Training Region-Based Object Detectors with Online Hard Example Mining // Proc of the IEEE Conference on Computer Vision and Pattern Recognition. Washington, USA: IEEE, 2016: 761-769.
[14] HERMANS A, BEYER L, LEIBE B.In Defense of the Triplet Loss for Person Reidentification[C/OL]. [2021-05-06].https://arxiv.org/pdf/1703.07737.pdf.
[15] XUAN H, STYLIANOU A, LIU X T, et al. Hard Negative Examples Are Hard, But Useful // Proc of the European Conference on Computer Vision. Berlin, German: Springer, 2020: 126-142.
[16] GEIGER A, LENZ P, STILLER C, et al. Vision Meets Robotics: The KITTI Dataset. International Journal of Robotics Research, 2013, 32(11): 1231-1237.
[17] CARLEVARIS-BIANCO N, USHANI A K, EUSTICE R M.University of Michigan North Campus Long-Term Vision and Lidar Dataset. International Journal of Robotics Research, 2016, 35(9):1023-1035.
[18] ZHANG J, SINGH S.LOAM: Lidar Odometry and Mapping in Real-time[C/OL]. [2021-05-06].https://frc.ri.cmu.edu/~zhangji/publications/RSS_2014.pdf.