﻿ 旋翼无人机单目视觉障碍物径向光流检测法
 文章快速检索 高级检索

1. 西安测绘总站, 陕西 西安 710000;
2. 信息工程大学导航与空天目标工程学院, 河南 郑州 450000

Monocular Vision Obstacle Detection Method Based on Radial Optical Flow for Rotor UAV
ZHANG Xiaodong1,2, HAO Xiangyang2, SUN Guopeng2, XU Yali2
1. Xi'an Division of Surveying and Mapping, Xi'an 710000, China;
2. Information Engineering University, School of Navigation and Aerospace Engineering, Zhengzhou 450000, China
Foundation support: Information Engineering University "2110 Project" Construction Project (No. 510087)
First author: ZHANG Xiaodong(1991-), male, master, majors in visual navigation and visual inspection.E-mail:1228024443@qq.com
Corresponding author: HAO Xiangyang
Abstract: To solve the problem of traditional Pyramid LK optical flow algorithm's poor accuracy and adaptability for rotor UAV to detect obstacle in complex outdoor environment, a monocular autonomous real-time obstacle detection method based on radial optical flow is proposed. In the optical flow, the radial optical flow is computed by fusing Pyramid LK optical flow with tangential optical flow, and a new obstacles decision strategy to detect obstacles based on the radial optical flow is put forward. Experimental results show that without increasing the complexity of algorithm, the proposed method can get a higher accuracy and better adaptability than traditional Pyramid LK algorithm, which can meet the requirements of UAV autonomous obstacle avoidance.
Key words: rotor UAV     monocular vision     obstacle detection     radial optical flow     Pyramid LK optical flow     tangential optical flow

1 径向光流原理及计算

1.1 金字塔LK光流

LK光流算法基于以下3个假设，相邻两幅影像之间满足：① 亮度恒定；② 运动是“小运动”；③ 空间一致，临近点有相似运动，保持相邻[19]

(1)

(2)

(3)

(4)

(5)

(6)

LK方法要求必须满足小运动，亮度不变以及区域一致，但无人机运动较快使图像中近距离物体产生的光流变化较大，不能满足假设条件，导致最终求得的光流值有较大的偏差，因此本文采用图像金字塔对光流进行从粗到精的估计，如图 1所示。在图像金字塔的最高层计算光流，用得到的运动估计结果作为下一层金字塔的起始点，重复这个过程至金字塔最底层，这样就能将不满足运动假设的可能性降到最低[20]

 图 1 金字塔LK光流原理 Fig. 1 Principle of Pyramid LK optical flow

1.2 切向光流

(7)
 图 2 切向光流示意图 Fig. 2 Tangential flow diagram

(8)

(9)
1.3 光流融合计算径向光流

(10)
 图 3 光流示意图 Fig. 3 Diagram of the optical flow

2 径向光流检测障碍物

2.1 光流的原理与分析

 图 4 针孔摄像机几何模型 Fig. 4 Geometric model of pinhole camera

 图 5 两帧图像的成像模型 Fig. 5 The imaging model of two-frame image

(11)

(12)

2.2 障碍物检测

 图 6 障碍物检测模型 Fig. 6 Obstacle detection model

(13)

(14)

(1) 通过金字塔LK光流法求解两帧图像上同名点的光流向量集合

(15)

(2) 计算两帧图像的切向光流vq

(3) 求解金字塔LK光流与切向光流的矢量差，得到径向光流的向量集合

(16)

(4) 计算每个径向光流的大小

(17)

(5) 计算前一帧图像上的特征点到图像中心点的距离

(18)

(6) 将径向光流中|vji|＞0.3h1i的奇异点从向量集合中剔除。

(7) 根据已知的危险距离L、基线长度D设定检测阈值W

(19)

(8) 比较径向光流与检测阈值，当|vji|＞Wi时，认为障碍物进入无人机的危险区域。

3 试验

3.1 半实物仿真试验

 图 7 高精度视觉导航工控平台 Fig. 7 High-precision visual navigation industrial control platform

 图 8 金字塔LK光流检测障碍物 Fig. 8 Pyramid LK optical flow detect obstacle

 图 9 径向光流检测障碍物 Fig. 9 Radial optical flow detect obstacle

 图 10 障碍物检测结果 Fig. 10 Obstacle detection results

3.2 无人机实际飞行试验

 图 11 障碍物检测结果 Fig. 11 Obstacle detection results

 图 12 障碍物检测结果 Fig. 12 Obstacle detection results

4 结论

(1) 径向光流检测障碍物的准确性较高，当被检测物体进入无人机危险区域时可以准确检测到障碍物。

(2) 径向光流能够规避无人机切向移动引入的误差，可用于室外环境下旋翼无人机自主避障。

﻿

 [1] REINEMAN B D, LENAIN L, STATOM N M, et al. Development and Testing of Instrumentation for UAV-based Flux Measurements within Terrestrial and Marine Atmospheric Boundary Layers[J]. Journal of Atmospheric and Oceanic Technology, 2013, 30(7): 1295–2954. DOI:10.1175/JTECH-D-12-00176.1 [2] 马源. 基于双目立体视觉的深度感知技术研究[D]. 北京: 北京理工大学, 2015. MA Yuan. Research on Depth Perception Based on the Binocular Stereo Vision[D]. Beijing:Beijing Institute of Technology, 2015. http://cdmd.cnki.com.cn/Article/CDMD-10007-1015030756.htm [3] CHAO Haiyang, CAO Yongcan, CHEN Yangquan. Autopilots for Small Unmanned Aerial Vehicles:A Survey[J]. International Journal of Control, Automation and Systems, 2010, 8(1): 36–44. DOI:10.1007/s12555-010-0105-z [4] ZINGG S, SCARAMUZZA D, WEISS S, et al. MAV Navigation through Indoor Corridors Using Optical Flow[C]//Proceedings of the 2010 IEEE International Conference on Robotics and Automation (ICRA). Anchorage, AK:IEEE, 2010:3361-3368. [5] LIYANAGE D K, PERERA M U S. Optical Flow Based Obstacle Avoidance for the Visually Impaired[C]//Proceedings of the 2012 IEEE Business Engineering and Industrial Applications Colloquium (BEIAC). Kuala Lumpur:IEEE, 2012:284-289. [6] LIAU Y S, ZHANG Qun, LI Yanan, et al. Non-metric Navigation for Mobile Robot Using Optical Flow[C]//Proceedings of the 2012 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). Vilamoura:IEEE, 2012:4953-4958. [7] SRINIVASAN M V, MOORE R J D, THURROWGOOD S, et al. From Biology to Engineering:Insect Vision and Applications to Robotics[M]//BARTH F G, HUMPHREY J A C, SRINIVASAN M V. Frontiers in Sensing. Vienna:Springer, 2012:19-39. [8] 徐伟杰. 基于视觉的微小型无人直升机位姿估计与目标跟踪研究[D]. 杭州: 浙江大学, 2012. XU Weijie. Research on Vision-based Pose Estimation and Object Tracking for Miniature Unmanned Helicopter[D]. Hangzhou:Zhejiang University, 2012. http://cdmd.cnki.com.cn/Article/CDMD-10335-1013185874.htm [9] 赵海, 陈星池, 王家亮, 等. 基于四轴飞行器的单目视觉避障算法[J]. 光学精密工程, 2014, 22(8): 2232–2241. ZHAO Hai, CHEN Xingchi, WANG Jialiang, et al. Obstacle Avoidance Algorithm Based on Monocular Vision for Quad-rotor Helicopter[J]. Optics and Precision Engineering, 2014, 22(8): 2232–2241. [10] ERESEN A, İMAMOǦLU N, ÖNDER EFE M. Autonomous Quad Rotor Flight with Vision Based Obstacle Avoidance in Virtual Environment[J]. Expert Systems with Applications, 2012, 39(1): 894–905. DOI:10.1016/j.eswa.2011.07.087 [11] 易礼智. 基于机器视觉的避障智能小车系统研究[D]. 长沙: 中南大学, 2012. YI Lizhi. Research on Obstacle Avoidance Smart Car System Based on Machine Vision[D]. Changsha:Central South University, 2012. http://cdmd.cnki.com.cn/Article/CDMD-10533-1014148321.htm [12] 郑海华. 基于视觉的平地探测机器人避障研究[D]. 长沙: 中南大学, 2014. ZHENG Haihua. Research on Obstacle Avoidance of Flat Detection Robot Based on Vision[D]. Changsha:Central South University, 2014. http://cdmd.cnki.com.cn/Article/CDMD-10533-1014408359.htm [13] 刘正东, 高鹏, 杨静宇. 一种用于道路避障的双目视觉图像分割方法[J]. 计算机应用研究, 2005, 22(4): 249–251. LIU Zhengdong, GAO Peng, YANG Jingyu. Binocular Image Segmentation Algorithm for Obstacles Avoidance[J]. Application Research of Computers, 2005, 22(4): 249–251. [14] KENDOUL F, FANTONI A, NONAM A. Optic Flow-based Vision System for Autonomous 3D Localization and Control of Small Aerial Vehicles[J]. Robotics and Autonomous Systems, 2009, 57(6-7): 591–602. DOI:10.1016/j.robot.2009.02.001 [15] GRIGORESCU S M, MACESANU G, COCIAS T T, et al. Robust Camera Pose and Scene Structure Analysis for Service Robotics[J]. Robotics and Autonomous Systems, 2011, 59(11): 889–909. [16] VO V, LY N. An Effective Approach for Human Actions Recognition Based on Optical Flow and Edge Features[C]//Proceedings of the 2012 International Conference on Control, Automation and Information Sciences (ICCAIS). Ho Chi Minh City:IEEE, 2012:24-29. [17] CHATTERJEE A, RAKSHIT A, SINGH N N. 基于视觉的自主机器人导航[M]. 连晓峰, 译. 北京: 机械工业出版社, 2014. CHATTERJEE A, RAKSHIT A, SINGH N N. Vision Based Autonomous Robot Navigation[M]. LIAN Xiaofeng, trans. Beijing:Machinery Industry Press, 2014. [18] 张正鹏, 江万寿, 张靖. 自适应运动结构特征的车载全景序列影像匹配方法[J]. 测绘学报, 2015, 44(10): 1132–1141. ZHANG Zhengpeng, JIANG Wanshou, ZHANG Jing. An Matching Method for Vehicle-borne Panoramic Image Sequence Based on Adaptive Structure from Motion Feature[J]. Acta Geodaetica et Cartographica Sinica, 2015, 44(10): 1132–1141. DOI:10.11947/j.AGCS.2015.20140622 [19] BRADSKI G, KAEHLER A. 学习OpenCV[M]. 于仕琪, 刘瑞祯, 译. 北京: 清华大学出版社, 2009: 358-366. BRADSKI G, KAEHLER A. Learning OpenCV[M]. YU Shiqi, LIU Ruizhen, trans. Beijing:Tsinghua University Press, 2009:358-366. [20] 李佳田, 李显凯, 李应芸, 等. 航空图像光流场的逆向金字塔计算方法[J]. 测绘学报, 2016, 45(9): 1059–1064. LI Jiatian, LI Xiankai, LI Yingyun, et al. A Backward Pyramid Oriented Optical Flow Field Computing Method for Aerial Image[J]. Acta Geodaetica et Cartographica Sinica, 2016, 45(9): 1059–1064. DOI:10.11947/j.AGCS.2016.20150367 [21] HARTLEY R, ZISSERMAN A. Multiple View Geometry in Computer Vision[M]. Cambridge: Cambridge University Press, 2004. [22] EBERLI D, SCARAMUZZA D, WEISS S, et al. Vision Based Position Control for MAVs Using One Single Circular Landmark[J]. Journal of Intelligent & Robotic System, 2011, 61(1-4): 495–512. [23] HARTLEY R, ZISSERMAN A. 计算机视觉中的多视图几何[M]. 韦穗, 杨尚骏, 章权兵, 译. 合肥: 安徽大学出版社, 2002. HARTLEY R, ZISSERMAN A. Multiple View Geometry in Computer Vision[M]. WEI Hui, YANG Shangjun, ZHANG Quanbing, trans. Hefei:Anhui University Press, 2002. [24] 崔燕茹. 基于双目视觉的障碍物识别与重建[D]. 南昌: 南昌航空大学, 2012. CUI Yanru. Obstacle Recognition and Reconstruction Based on Binocular Vision[D]. Nanchang:Nanchang Hangkong University, 2012. http://cdmd.cnki.com.cn/Article/CDMD-10406-1012032800.htm [25] STEFANOSKI N, WANG O, LANG M, et al. Automatic View Synthesis by Image-domain-warping[J]. IEEE Transactions on Image Processing, 2013, 22(9): 3329–3341. DOI:10.1109/TIP.2013.2264817 [26] 夏军营, 徐小泉, 熊九龙. 利用平行透视投影模型的位姿迭代估计[J]. 光学精密工程, 2012, 20(6): 1342–1349. XIA Junying, XU Xiaoquan, XIONG Jiulong. Iterative Pose Estimation Using Paraperspective Camera Model[J]. Optics and Precision Engineering, 2012, 20(6): 1342–1349. [27] 戴碧霞. 基于光流的微小型飞行器室内避障方法研究[D]. 成都: 电子科技大学, 2015. DAI Bixia. Research on Obstacle Avoidance Method of MAV in Indoor Environment Based on Optical Flow[D]. Chengdu:University of Electronic Science and Technology of China, 2015. http://cdmd.cnki.com.cn/Article/CDMD-10614-1015708658.htm
http://dx.doi.org/10.11947/j.AGCS.2017.20160510

0

#### 文章信息

ZHANG Xiaodong, HAO Xiangyang, SUN Guopeng, XU Yali

Monocular Vision Obstacle Detection Method Based on Radial Optical Flow for Rotor UAV

Acta Geodaetica et Cartographica Sinica, 2017, 46(9): 1107-1115
http://dx.doi.org/10.11947/j.AGCS.2017.20160510