无人机室内视觉/惯导组合导航方法

王亭亭; 蔡志浩; 王英勋

doi:10.13700/j.bh.1001-5965.2016.0965

无人机室内视觉/惯导组合导航方法

doi: 10.13700/j.bh.1001-5965.2016.0965

王亭亭¹,
蔡志浩^{1, 2, ,},
王英勋^{1, 2}

1.
北京航空航天大学自动化科学与电气工程学院, 北京 100083
2.
北京航空航天大学飞行器控制一体化技术国防科技重点实验室, 北京 100083

基金项目:

航空科学基金 20135851043

详细信息

作者简介:
王亭亭  女, 硕士研究生。主要研究方向:机器视觉、无人机视觉导航

蔡志浩  男, 副教授, 硕士生导师。主要研究方向:无人机自主控制与导航、多机协同与训练

王英勋  男, 研究员, 博士生导师。主要研究方向:系统设计、自主控制与模拟训练

通讯作者:
蔡志浩, E-mail: czh@buaa.edu.cn

中图分类号: TP242.6;V249.32
计量
- 文章访问数: 1620
- HTML全文浏览量: 92
- PDF下载量: 889
- 被引次数: 0
出版历程
- 收稿日期: 2016-12-23
- 录用日期: 2017-02-06
- 网络出版日期: 2018-01-20

Integrated vision/inertial navigation method of UAVs in indoor environment

WANG Tingting¹,
CAI Zhihao^{1, 2
, ,},
WANG Yingxun^{1, 2}

1.
School of Automation Science and Electrical Engineering, Beijing University of Aeronautics and Astronautics, Beijing 100083, China
2.
Aircraft Control Integration National Defense Key Laboratory, Beijing University of Aeronautics and Astronautics, Beijing 100083, China

Funds:

Aeronautical Science Foundation of China 20135851043

More Information

Corresponding author: CAI Zhihao, E-mail: czh@buaa.edu.cn

摘要

摘要:
针对室内无卫星定位下的无人机自主导航问题，提出了一种融合惯导、光流和视觉里程计的组合导航方法。在速度估计上，采用基于ORB特征的光流法，该方法可以实时地估计出无人机的三轴线速度信息。方法采用基于特征点的稀疏光流，对金字塔Lucas-Kanade光流法进行了改进，采用前后双向追踪和随机采样一致的方法提高特征点追踪精度。在位置估计上，采用视觉/惯导融合的视觉里程计，以人工图标法为主，融合视觉光流信息和惯导数据实现无人机定位。通过与运动捕捉系统的定位信息、Guidance和PX4Flow导航模块的测速信息进行对比，以及实际的飞行测试，验证本文方法的可行性。
- 无人机 /
- 视觉导航 /
- 光流 /
- ORB特征 /
- 多传感器融合
Abstract:
A new integrated navigation method based on inertial sensor, optical flow and visual odometry is proposed for self-navigation indoor in GPS-denied environment. An ORB optical flow based method is also proposed for estimating real-time three-axis velocity of the UAV. The algorithm improves the traditional pyramid Lucas-Kanade method using sparse optical flow based on feature points. The tracking of feature points is made more accurate by applying forward-backward tracking and random sampling consensus strategies. For position estimation, a visual odometry method with integrated vision/inertial navigation is adopted, which uses the artificial icon method, visual optical flow information and inertial navigation data. Finally, the velocity and position estimations from the proposed method are validated via actual flight test and via comparison with velocity measurement information from a PX4Flow module and a Guidance module and with locating information from movement capture system.
- UAV /
- vision navigation /
- optical flow /
- ORB features /
- multi-sensor fusion

HTML全文

图 1 前后双向追踪示意图

Figure 1. Schematic of forward-backward tracking

下载: 全尺寸图片幻灯片

图 2 采用RANSAC算法对特征点对滤波

Figure 2. Filtering features by RANSAC algorithm

下载: 全尺寸图片幻灯片

图 3 相机运动时空间点成像变化

Figure 3. Point-mapping changing caused by moving camera

下载: 全尺寸图片幻灯片

图 4 姿态变化时测速对比

Figure 4. Velocity comparison with changing attitude

下载: 全尺寸图片幻灯片

图 5 图标法里程计算法基本流程

Figure 5. Algorithm flowchart of icon odometry

下载: 全尺寸图片幻灯片

图 6 图像滤波与边缘提取

Figure 6. Image filtering and edge extraction

下载: 全尺寸图片幻灯片

图 7 图标法里程计定位

Figure 7. Positioning based on icon odometry

下载: 全尺寸图片幻灯片

图 8 定位测试

Figure 8. Localization test

下载: 全尺寸图片幻灯片

图 9 三维位置估计与位置误差

Figure 9. Localization estimation and positioning error in 3D-directions

下载: 全尺寸图片幻灯片

图 10 三维/二维运动轨迹

Figure 10. Three/two-dimensional motion trajectory

下载: 全尺寸图片幻灯片

图 11 三维速度估计与速度误差

Figure 11. Velocity estimation and velocity error in 3D-directions

下载: 全尺寸图片幻灯片

图 12 无人机平台及系统结构

Figure 12. UAV platform and system structure

下载: 全尺寸图片幻灯片

图 13 飞行视频截图

Figure 13. Screenshots of flight video

下载: 全尺寸图片幻灯片

图 14 手控飞行速度测试

Figure 14. Flight velocity test via manual control

下载: 全尺寸图片幻灯片

图 15 位置估计

Figure 15. Position estimation

下载: 全尺寸图片幻灯片

图 16 速度估计与飞机姿态

Figure 16. Velocity estimation and aircraft attitude

下载: 全尺寸图片幻灯片

表 1 SIFT、SURF和ORB特征提取时间开销对比

Table 1. Comparison of time consumption of feature extraction among SIFT, SURF and ORB

方法	特征点数	时间/ms
SIFT^[14]	171	18.89
	253	18.99
	234	19.66
SURF^[15]	86	12.44
	254	16.21
	187	13.86
ORB	168	2.82
	299	4.11
	251	4.7

下载: 导出CSV

参考文献(18)

[1]	SHEN S J.Autonomous navigation in complex indoor and outdoor environments with micro aerial vehicles[D].Philadelphia:University of Pennsylvania, 2014.
[2]	吴琦, 蔡志浩, 王英勋.用于无人机室内导航的光流与地标融合方法[J].控制理论与应用, 2015, 32(11):1511-1517. http://www.oalib.com/paper/4746150 WU Q, CAI Z H, WANG Y X.Optical flow and landmark fusion method for UAV indoor navization[J].Control Theory & Applications, 2015, 32(11):1511-1517(in Chinese). http://www.oalib.com/paper/4746150
[3]	LI P, LAMBERT A.A monocular odometer for a quadrotor using a homogra-phy model and inertial cues[C]//IEEE Conference on Robotics and Biomimetics.Piscataway, NJ:IEEE Press, 2015:570-575.
[4]	叶长春. IARC第7代任务中定位与目标跟踪方法研究[D]. 杭州: 浙江大学, 2016. YE C C.Research on localization and object tracking for the IARC mission7[D].Hangzhou:Zhejiang University, 2016(in Chinese).
[5]	MUR-ARTAL R, MONTIEL J M M, TARDOS J D.ORB-SLAM:A versatile and accurate monocular SLAM system[J].IEEE Transactions on Robotics, 2015, 31(5):1147-1163. doi: 10.1109/TRO.2015.2463671
[6]	LEUTENEGGER S, FURGALE P, RABAUD V, et al.Keyframe-based visual-inertial SLAM using nonlinear optimization[C]//Robotics:Science and Systems, 2013:789-795.
[7]	CHAO H, GU Y, GROSS J, et al.A comparative study of optical flow and traditional sensors in UAV navigation[C]//American Control Conference(ACC).Piscataway, NJ:IEEE Press, 2013:3858-3863.
[8]	MAMMARELLA M, CAMPA G, FRAVOLINI M L, et al.Comparing optical flow algorithms using 6-dof motion of real-world rigid objects[J].IEEE Transactions on, Systems, Man, and Cybernetics, Part C:Applications and Reviews, 2012, 42(6):1752-1762. doi: 10.1109/TSMCC.2012.2218806
[9]	DJI Innovations.PHANTOM 4 user's manual V1.2[EB/OL].(2016-12-23)https://dl.djicdn.com/downloads/phantom_4/cn/Phantom_4_User_Manual_cn_v1.2_160328.pdf.
[10]	Hover Camera 2016[EB/OL].(2016-12-23)http://gethover.com.
[11]	RUBLEE E, RABAUD V, KONOLIGE K, et al.ORB:An efficient alternative to SIFT or SURF[J].Proceedings, 2011, 58(11):2564-2571. https://www.willowgarage.com/sites/default/files/orb_final.pdf
[12]	ROSTEN E, DRUMMOND T.Machine learning for high-speed corner detection[C]//European Conference on Computer Vision.Berlin:Springer-Verlag, 2006:430-443.
[13]	CALONDER M, LEPETIT V, STRECHA C, et al.BRIEF:Binary robust independent elementary features[C]//European Conference on Computer Vision.Berlin:Springer-Verlag, 2010:778-792.
[14]	LOWE D G.Distinctive image features from scale-invariant keypoints[J].International Journal of Computer Vision, 2004, 60(2):91-110. doi: 10.1023/B:VISI.0000029664.99615.94
[15]	BAY H, TUYTELAARS T, GOOL L V.SURF:Speeded up robust features[J].Computer Vision & Image Understanding, 2006, 110(3):404-417. http://www.vision.ee.ethz.ch/~surf/eccv06.pdf
[16]	CHUM O, MATAS J, KITTLER J.Locally optimized RANSAC[J].Lecture Notes in Computer Science, 2003, 2781:236-243. doi: 10.1007/b12010
[17]	HONEGGER D, MEIER L, TANSKANEN P, et al.An open source and open hardware embedded metric optical flow cmos camera for indoor and outdoor applications[C]//International Conference on Robotics and Automation.Piscataway, NJ:IEEE Press, 2013:1736-1741.
[18]	HARTLEY R I, ZISSERMAN A.Multi-view geometry in computer vision[M].Cambrige:Cambridge University Press, 2004:239-247.