首页 > 论文 > 激光与光电子学进展 > 56卷 > 18期(pp:181501--1)

融合Gist特征与卷积自编码的闭环检测算法

Loop Closure Detection Algorithm Based on Convolutional Autoencoder Fused with Gist Feature

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

摘要

闭环检测算法可消除视觉同时定位与建图(VSLAM)系统的累计误差,并对构建全局一致性地图有重要作用。针对现有传统闭环检测算法在视角与场景外观变化下准确率与稳健性降低,及部分基于深度学习方法特征提取与闭环识别实时性不佳的问题,设计了一种融合Gist特征与卷积自编码的闭环检测算法,将Gist特征作为卷积自编码网络重构目标,可增强模型在外观变化下的场景特征表达能力;同时通过透视变换构造视角变化训练图像对,以提升模型在视角变化下闭环检测的准确率与稳健性。所设计的模型较精简,可实现实时关键帧特征提取与闭环检测。在Gardens Point与Nordland数据集的实验结果表明,相较于传统视觉词袋模型(BoVW)、Gist算法及现有部分深度学习方法,本文算法可以达到更高的准确率和稳健性。

Abstract

Loop closure detection algorithm is essential for the visual simultaneous localization and mapping (VSLAM) systems to reduce accumulative error and build a globally consistent map. When detecting loops under the change of viewpoint and scene appearance, the precision and robustness of traditional loop closure detection algorithms decline and some algorithms based on deep learning are difficult to extract features and perform loop closure detection in real time. To overcome these problems, we propose a novel loop closure detection algorithm based on convolutional autoencoder fused with Gist feature, forcing the encoder to reconstruct the Gist feature to enhance the expressive ability of the model when the scene appearance changes. In the same time, we warp images with randomized projective transformations to make the training pairs to improve the precision and robustness of the model when the viewpoint changes. Our model is relatively lightweight which is capable of extracting keyframe features and detecting loops in real time. The results of experiments on Gardens Point and Nordland datasets show that our model can achieve better precision and robustness compared with traditional methods, like bag of visual word (BoVW), Gist, and some other methods based on deep learning.

Newport宣传-MKS新实验室计划
补充资料

DOI:10.3788/LOP56.181501

所属栏目:机器视觉

基金项目:微系统技术国防科技重点实验室基金;

收稿日期:2019-01-12

修改稿日期:2019-04-09

网络出版日期:2019-09-01

作者单位    点击查看

邱晨力:中国科学院上海微系统与信息技术研究所微系统技术重点实验室, 上海 201800中国科学院大学, 北京 100049
黄东振:中国科学院上海微系统与信息技术研究所微系统技术重点实验室, 上海 201800中国科学院大学, 北京 100049
刘华巍:中国科学院上海微系统与信息技术研究所微系统技术重点实验室, 上海 201800
袁晓兵:中国科学院上海微系统与信息技术研究所微系统技术重点实验室, 上海 201800
李宝清:中国科学院上海微系统与信息技术研究所微系统技术重点实验室, 上海 201800

联系人作者:李宝清(sinoiot@mail.sim.ac.cn)

备注:微系统技术国防科技重点实验室基金;

【1】Lin Z L, Zhang G L, Yao E L et al. Stereo visual odometry based on motion object detection in the dynamic scene. Acta Optica Sinica. 37(11), (2017).
林志林, 张国良, 姚二亮 等. 动态场景下基于运动物体检测的立体视觉里程计. 光学学报. 37(11), (2017).

【2】Wang K, Jia S M, Li X Z et al. Mobile robot monocular visual odometry algorithm based on ground features. Acta Optica Sinica. 35(5), (2015).
王可, 贾松敏, 李秀智 等. 基于地面特征的移动机器人单目视觉里程计算法. 光学学报. 35(5), (2015).

【3】Galvez-López D and Tardos J D. Bags of binary words for fast place recognition in image sequences. IEEE Transactions on Robotics. 28(5), 1188-1197(2012).

【4】Bay H. Tuytelaars T, van Gool L. SURF: speeded up robust features. //Leonardis A, Bischof H, Pinz A. Computer vision-ECCV 2006. Lecture notes in computer science. Berlin Heidelberg: Springer. 3951, 404-417(2006).

【5】Rublee E, Rabaud V, Konolige K et al. ORB: an efficient alternative to SIFT or SURF. [C]//2011 International Conference on Computer Vision, November 6-13, 2011, Barcelona, Spain. New York: IEEE. 2564-2571(2011).

【6】Cummins M and Newman P. FAB-MAP: probabilistic localization and mapping in the space of appearance. The International Journal of Robotics Research. 27(6), 647-665(2008).

【7】Mur-Artal R. Montiel J M M, Tardós J D. ORB-SLAM: a versatile and accurate monocular SLAM system. IEEE Transactions on Robotics. 31(5), 1147-1163(2015).

【8】Oliva A and Torralba A. Building the gist of a scene: the role of global image features in recognition. Progress in Brain Research. 155, 23-36(2006).

【9】Liu Y and Zhang H. Visual loop closure detection with a compact image descriptor. [C]//2012 IEEE/RSJ International Conference on Intelligent Robots and Systems, October 7-12, 2012, Vilamoura, Portugal. New York: IEEE. 1051-1056(2012).

【10】Hou Y, Zhang H and Zhou S L. Convolutional neural network-based image representation for visual loop closure detection. [C]//2015 IEEE International Conference on Information and Automation, August 8-10, 2015, Lijiang, China. New York: IEEE. 2238-2245(2015).

【11】Zhou B L, Lapedriza A, Xiao J X et al. Learning deep features for scene recognition using places database[C]//Proceedings of the 27th International Conference on Neural Information Processing Systems, December 8-13, 2014, Montreal, Canada. 487-495(2014).

【12】Sünderhauf N, Shirazi S, Dayoub F et al. On the performance of ConvNet features for place recognition. [C]//2015 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), September 28-October 2, 2015, Hamburg, Germany. New York: IEEE. 4297-4304(2015).

【13】Gao X and Zhang T. Loop closure detection for visual SLAM systems using deep neural networks. [C]//2015 34th Chinese Control Conference (CCC), July 28-30, 2015, Hangzhou, China. New York: IEEE. 5851-5856(2015).

【14】Gao X and Zhang T. Unsupervised learning to detect loops using deep neural networks for visual SLAM system. Autonomous Robots. 41(1), 1-18(2017).

【15】Lopez-Antequera M, Gomez-Ojeda R, Petkov N et al. Appearance-invariant place recognition by discriminatively training a convolutional neural network. Pattern Recognition Letters. 92, 89-95(2017).

【16】Merrill N. -05-24)[2018-12-29]. Huang G Q. Lightweight unsupervised deep loop closure. org/abs/1805, (2018).

【17】Bao Z Q, Li A H, Cui Z G et al. Loop closure detection algorithm based on multi-level convolutional neural network features. Laser & Optoelectronics Progress. 55(11), (2018).
鲍振强, 李艾华, 崔智高 等. 融合多层次卷积神经网络特征的闭环检测算法. 激光与光电子学进展. 55(11), (2018).

【18】Redmon J and Farhadi A. YOLO9000: better, faster, stronger. [C]//2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), July 21-26, 2017, Honolulu, HI. New York: IEEE. 6517-6525(2017).

【19】Ioffe S. -03-02)[2018-12-29]. Szegedy C. Batch normalization: accelerating deep network training by reducing internal covariate shift. org/abs/1502, (2015).

【20】Glorot X, Bordes A and Bengio Y. Deep sparse rectifier neural networks. [C]//14th International Conference on Artificial Intelligence and Statistics, April 11-13, 2011, Fort Lauderdale, USA. Cambridge: PMLR. 15, 315-323(2011).

【21】He K M and Sun J. Convolutional neural networks at constrained time cost. [C]//2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), June 7-12, 2015, Boston, MA, USA. New York: IEEE. 5353-5360(2015).

引用该论文

Chenli Qiu,Dongzhen Huang,Huawei Liu,Xiaobing Yuan,Baoqing Li. Loop Closure Detection Algorithm Based on Convolutional Autoencoder Fused with Gist Feature[J]. Laser & Optoelectronics Progress, 2019, 56(18): 181501

邱晨力,黄东振,刘华巍,袁晓兵,李宝清. 融合Gist特征与卷积自编码的闭环检测算法[J]. 激光与光电子学进展, 2019, 56(18): 181501

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