首页 > 论文 > 激光与光电子学进展 > 55卷 > 11期(pp:111507--1)

融合多层次卷积神经网络特征的闭环检测算法

Loop Closure Detection Algorithm Based On Multi-Level Convolutional Neural Network Features

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

摘要

在外部环境和图像视角变化的情况下,传统视觉闭环检测算法的精度和稳健性变得很差。为此,提出一种融合多层次卷积神经网络特征的闭环检测算法。高层次的卷积特征包含较多的语义信息,可以应对图像视角的变化;中等层次卷积特征包含更多的几何空间信息,对光照等变化具有更好的稳健性。通过充分利用中高等层次卷积特征的特性进行组合式相似性度量,提高了闭环检测的精度与稳健性。由于卷积特征向量的维度特别大,因此,首先对卷积特征向量进行降维处理。在Gardens Point数据集上的实验结果证明,利用多层次卷积特征的图像匹配检测效果好于其他单一层。针对不同时刻所拍摄图像中的动态干扰因素,进一步提出图像动态干扰语义滤波机制,利用过滤掉动态干扰的图像进行匹配,在Tokyo24/7数据集上的实验证实了此方法的可行性和有效性。

Abstract

In the cases of appearance changes and viewpoint changes, the accuracy and robustness of traditional visual loop closure detection algorithms become very poor.To overcome this problem, we propose a loop closure detection algorithm, which utilizes the features of multi-level convolutional neural networks. The high-level convolution features contain much semantic information and can cope with viewpoint changes. The medium-level convolutional features contain more geometry and spatial information, which is more robust to lighting changes. Therefore, the accuracy and robustness of loop closure detection is improved by taking full advantage of the characteristics of the middle and high levels convolutional features and modular similarity measures. However, the convolutional feature vectors have a particularly large dimension, so the convolutional feature vectors are firstly dimension-reduced. The experimental results on the Gardens Point dataset show that the image matching detection effect is better by using multi-level convolutional features than by other single layers. In addition, for the dynamic interference factors in the images captured at different moments, a dynamic interference semantic filtering mechanism is further proposed. The filtered images are used to perform the matching. The experiments on the Tokyo 24/7 dataset prove the feasibility and effectiveness of this method.

Newport宣传-MKS新实验室计划
补充资料

中图分类号:TP242

DOI:10.3788/lop55.111507

所属栏目:机器视觉

基金项目:国家自然科学基金(61501470)、陕西省重点研发计划(2017GY-075)

收稿日期:2018-04-21

修改稿日期:2018-05-21

网络出版日期:2018-06-06

作者单位    点击查看

鲍振强:火箭军工程大学作战保障学院, 陕西 西安 710025
李艾华:火箭军工程大学作战保障学院, 陕西 西安 710025
崔智高:火箭军工程大学作战保障学院, 陕西 西安 710025
苏延召:火箭军工程大学作战保障学院, 陕西 西安 710025
郑勇:火箭军工程大学作战保障学院, 陕西 西安 710025

联系人作者:鲍振强(bzhenqiang@163.com)

【1】Lin Z L, Zhang G L, Yao E L, et al. Stereo visual odometry based on motion object detection in the dynamic scene[J]. Acta Optica Sinica, 2017, 37(11): 1115001.
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[J]. 光学学报, 2017, 37(11): 1115001.

【2】Zhang X, Jin Y X, Xue D. Image matching algorithm based on SICA-SIFT and particle swarm optimization[J]. Laser & Optoelectronics Progress, 2017, 54(9): 091002.
张鑫, 靳雁霞, 薛丹. SICA-SIFT和粒子群优化的图像匹配算法[J]. 激光与光电子学进展, 2017, 54(9): 091002.

【3】Shan B H, Huo X Y, Liu Y. A stereovision measurement method using epipolar constraint to correct digital image correlation matching[J]. Chinese Journal of Lasers, 2017, 44(8): 0804003.
单宝华, 霍晓洋, 刘洋. 一种极线约束修正数字图像相关匹配的立体视觉测量方法[J]. 中国激光, 2017, 44(8): 0804003.

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

【5】Bay H, Tuytelaars T, van Gool L. SURF: speeded up robust features[C]∥European Conference on Computer Vision, 2006: 404-417.

【6】Rublee E, Rabaud V, Konolige K, et al. ORB: an efficient alternative to SIFT or SURF[C]∥International Conference on Computer Vision, 2011: 2564-2571.

【7】Oliva A, Torralba A. Modeling the shape of the scene: a holistic representation of the spatial envelope [J]. International Journal of Computer Vision, 2001, 42(3): 145-175.

【8】Szegedy C, Liu W, Jia Y Q, et al. Going deeper with convolutions[C]∥IEEE Conference on Computer Vision and Pattern Recognition, 2015: 1-9.

【9】Zhou B, Lapedriza A, Xiao J, et al. Learning deep features for scene recognition using places database[C]∥Annual Conference on Neural Information Processing Systems, 2014: 487-495.

【10】Chen Z T, Lam O, Jacobson A, et al. Convolutional neural network-based place recognition[C]∥Computer Vision and Pattern Recognition, 2014.

【11】Hou Y, Zhang H, Zhou S L. Convolutional neural network-based image representation for visual loop closure detection[C]∥IEEE International Conference on Information and Automation, 2015: 2238-2245.

【12】Simonyan K, Zisserman A. Very deep convolutional networks for large-scale image recognition[C]∥International Conference on Learning Representations, 2015.

【13】Sünderhauf N, Shirazi S, Dayoub F, et al. On the performance of ConvNet features for place recognition[C]∥IEEE/RSJ International Conference on Intelligent Robots and Systems, 2015: 4297-4304.

【14】Charikar M S. Similarity estimation techniques from rounding algorithms[C]∥34th Annual ACM Symposium on Theory of Computing, 2002: 380-388.

【15】Redmon J, Farhadi A. YOLO9000: better, faster, stronger[C]∥IEEE Conference on Computer Vision and Pattern Recognition, 2017: 6517-6525.

【16】Torii A, Arandjelovi R, Sivic J, et al. 24/7 place recognition by view synthesis[C]∥IEEE Conference on Computer Vision and Pattern Recognition, 2015: 1808-1817.

引用该论文

Bao Zhenqiang,Li Aihua,Cui Zhigao,Su Yanzhao,Zheng Yong. Loop Closure Detection Algorithm Based On Multi-Level Convolutional Neural Network Features[J]. Laser & Optoelectronics Progress, 2018, 55(11): 111507

鲍振强,李艾华,崔智高,苏延召,郑勇. 融合多层次卷积神经网络特征的闭环检测算法[J]. 激光与光电子学进展, 2018, 55(11): 111507

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