一种基于曼哈顿世界假说下平面特征的RGB-D视觉室内定位方案

蒋育豪; 陈蕾

doi:10.3969/j.issn.1000-5641.2019.06.010

一种基于曼哈顿世界假说下平面特征的RGB-D视觉室内定位方案

doi: 10.3969/j.issn.1000-5641.2019.06.010

蒋育豪,
陈蕾^,

华东师范大学计算机科学与技术系, 上海 200062

详细信息

作者简介:
蒋育豪, 男, 硕士研究生, 研究方向为视觉定位.E-mail:krovkov@163.com

通讯作者:
陈蕾, 女, 副教授, 研究方向为室内定位.E-mail:lchen@cs.ecnu.edu.cn

中图分类号: TP399
计量
- 文章访问数: 133
- HTML全文浏览量: 48
- PDF下载量: 0
- 被引次数: 0
出版历程
- 收稿日期: 2019-01-11
- 刊出日期: 2019-11-25

A plane-based localization scheme using RGB-D sensor for the Manhattan World assumption

JIANG Yu-hao,
CHEN Lei^,

Department of Computer Science and Technology, East China Normal University, Shanghai 200062, China

摘要

摘要: 将曼哈顿世界假说（Manhattan World assumption，MW）引入室内定位问题，提出了一种改进的基于RGB-D视觉与平面特征的室内定位方案，不仅能有效提高场景匹配的成功率，还可简化未知场景下的定位问题，提高定位效率和实时性，可用于对同步定位与建图SLAM（SimultaneousLocalization and Mapping）系统的扩展.创新点主要体现在：针对解释树匹配的时间开销随特征数指数级上升的问题，设计了根据曼哈顿帧的主方向进行分解的匹配方法；针对单条行进路径搜索效率有待提高的问题，提出了在初始位姿确定后采用4自由度的简化定位方案；针对单帧中遍历执行子图匹配耗时较长的问题，将小范围子图合并为大范围子图后进行匹配.实验结果表明，该方案相较已有的平面特征定位方法，能缩短成功定位所需的行进距离，并显著降低单条行进路径上的平均搜索耗时.
- 定位 /
- 同步定位与建图 /
- 曼哈顿世界假说
Abstract: Using the Manhattan World assumption with plane-based indoor localization, an indoor positioning scheme based on plane features in RGB-D vision is proposed, which can be used to extend SLAM (Simultaneous Localization and Mapping) systems. A matching process based on the main direction of the Manhattan Frame is designed to reduce the exponentially increasing time consumed. Simplified localization with 4 degrees of freedom is adopted after the initial pose determination for the problem of low efficiency during exploration. Small subgraphs in each frame are merged into one subgraph for matching to reduce the time consumed for repetitive subgraph matching. The proposed scheme not only effectively increases the success rate of scene matching, but also simplifies positioning in unknown scenes and improves positioning efficiency. Experimental results show that the method can achieve successful localization with shorter path lengths and reduce computational cost for real-time applications.
- localization /
- SLAM (Simultaneous Localization and Mapping) /
- Manhattan World assumption

HTML全文

图 1 曼哈顿世界系列假说及关系图

Fig. 1 Representation of the environment in point cloud, plane model, MMW, AW, and MW

下载: 全尺寸图片幻灯片

图 2 PbMap建图过程

Fig. 2 Construction of PbMap

下载: 全尺寸图片幻灯片

图 3 姿态角术语

Fig. 3 Terminology of Euler angles

下载: 全尺寸图片幻灯片

图 4 平面配准流程

Fig. 4 Flow chart for plane registration

下载: 全尺寸图片幻灯片

图 5 行进路径中的主方向确定

Fig. 5 Determination of main directions in subsequent trajectories

下载: 全尺寸图片幻灯片

图 6 子图匹配耗时与匹配对数量关系

Fig. 6 Subgraph matching time-plane matches

下载: 全尺寸图片幻灯片

表 1 初始定位结果

Tab. 1 Results form initial localization

方法	邻近阈值	子图内	特征匹	搜索耗时	搜索耗时	总耗时	所需路径
方法	$tp_p$/m	特征数	配对数	(子图)/s	(帧)/s	/s	长度/m
	1	13.65	9.49	0.004 046	0.007 384	0.016 45	4.756 3
PbMap	2	21.53	15.17	0.007 494	0.014 702	0.022 80	4.321 9
	4	33.25	22.85	0.016 564	0.029 190	0.040 53	4.010 7
本文	3.71	32.42	31.08	0.010 168		0.022 34	4.056 9

下载: 导出CSV

表 2 路径定位平均结果

Tab. 2 Average result of localization on the trajectory

方法	邻近阈值$tp_p$/m	最大定位耗时/s	平均定位耗时/s	定位精度误差/m	识别成功率/%
	1	0.021 62	0.016 45	0.111 516	91.5
PbMap	2	0.031 18	0.022 80	0.099 691	91.5
	4	0.063 79	0.040 53	0.098 855	92
本文	3.71	0.026 94	0.019 068	0.099 435	92

下载: 导出CSV

参考文献(17)

[1]	GLOCKER B, SHOTTON J, CRIMINISI A, et al. Real-time RGB-D camera relocalization via randomized ferns for keyframe encoding[J]. IEEE transactions on visualization and computer graphics, 2015, 21(5):571-583. doi: 10.1109/TVCG.2014.2360403
[2]	SE S, LOWE D G, LITTLE J J. Vision-based global localization and mapping for mobile robots[J]. IEEE Transactions on robotics, 2005, 21(3):364-375. doi: 10.1109/TRO.2004.839228
[3]	MUR-ARTAL R, TARDÓS J D. Orb-slam2:An open-source slam system for monocular, stereo, and rgb-d cameras[J]. IEEE Transactions on Robotics, 2017, 33(5):1255-1262. doi: 10.1109/TRO.2017.2705103
[4]	SALAS-MORENO R F, NEWCOMBE R A, STRASDAT H, et al. Slam++: Simultaneous localisation and mapping at the level of objects[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2013: 1352-1359.
[5]	SHI Y, XU K, NIEßNER M, et al. Planematch: Patch coplanarity prediction for robust rgb-d reconstruction[C]//Proceedings of the European Conference on Computer Vision (ECCV). 2018: 750-766.
[6]	FORSTNER W, KHOSHELHAM K. Efficient and accurate registration of point clouds with plane to plane correspondences[C]//Proceedings of the IEEE International Conference on Computer Vision. 2017: 2165-2173.
[7]	TAGUCHI Y, JIAN Y D, RAMALINGAM S, et al. Point-plane SLAM for hand-held 3D sensors[C]//2013 IEEE International Conference on Robotics and Automation. IEEE, 2013: 5182-5189.
[8]	FERNÁNDEZ-MORAL E, RIVES P, ARÉVALO V, et al. Scene structure registration for localization and mapping[J]. Robotics and Autonomous Systems, 2016, 75:649-660. doi: 10.1016/j.robot.2015.09.009
[9]	SALAS-MORENO R F, GLOCKEN B, KELLY P H J, et al. Dense planar SLAM[C]//2014 IEEE International Symposium on Mixed and Augmented Reality (ISMAR). IEEE, 2014: 157-164.
[10]	HSIAO M, WESTMAN E, ZHANG G, ET AL. Keyframe-based dense planar SLAM[C]//2017 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2017: 5110-5117.
[11]	MA L, KERL C, STÜCKLER J, et al. CPA-SLAM: Consistent plane-model alignment for direct RGB-D SLAM[C]//2016 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2016: 1285-1291.
[12]	CHO H G, YEON S, CHOI H, et al. Detection and compensation of degeneracy cases for IMU-kinect integrated continuous SLAM with plane features[J]. Sensors, 2018, 18(4):935(9pages). doi: 10.3390/s18040935
[13]	COUGHLAN J M, YUILLE A L. Manhattan world: Compass direction from a single image by bayesian inference[C]//Proceedings of the Seventh IEEE International Conference on Computer Vision. IEEE, 1999, 2: 941-947.
[14]	SCHINDLER G, DELLAERT F. Atlanta world: An expectation maximization framework for simultaneous lowlevel edge grouping and camera calibration in complex man-made environments[C]//Proceedings of the 2004 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'04). IEEE, 2004, 1: Ⅰ-Ⅰ.
[15]	STRAUB J, ROSMAN G, FREIFELD O, et al. A mixture of manhattan frames: Beyond the manhattan world[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. 2014: 3770-3777.
[16]	RUSU R B, Cousins S. 3D is here: Point Cloud Library (PCL)[C]//IEEE International Conference on Robotics & Automation. 2011: 1-4.
[17]	GRIMSON W E L, LOZANO-PEREZ T. Localizing overlapping parts by searching the interpretation tree[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 1987, 9(4):469-482. doi: 10.1109-TPAMI.1987.4767935/