基于视觉导航的无人机位姿控制与自主返航技术

发布时间：2018-06-15 04:49

本文选题：旋翼无人机 + 计算机视觉　；参考：《上海交通大学》2015年硕士论文

【摘要】：近年来,旋翼无人机因其优秀的机动性能在各个领域的应用都受到了高度重视。为了使操控更便利、控制更精准,根据视觉信息对无人机进行位置姿态控制一直是研究热点。自主返航也是无人机的一个重要功能,目前多旋翼无人机产品的飞控系统大多提供基于卫星导航信号的一键返航功能,但是在某些复杂环境中卫星导航信号并不可靠。本文的研究目的就是基于视觉信息对旋翼无人机的位置姿态进行精确控制,并实现自主返航功能。首先,本文根据无人机底部垂直向下的摄像头图像的光流与惯导器件融合得到的无人机估算速度,引入速度控制器以改善无人机位姿控制的效果。四旋翼机飞行实验证明提出的控制算法能更好地实现基于地标物的定点悬停、自主循迹,同时对速度信息进行积分可以得到比较准确的无人机的相对位移。在此基础之上,本文利用无人机前部摄像头实现自主返航功能。根据计算机视觉多视角几何理论,对相机当前图像与关键帧进行匹配可以解算两个视角的三维变换关系,从而控制无人机逼近关键帧位置并悬停。本文提出的鲁棒的自主返航技术方案结合了估算速度导航与图像匹配悬停:在去程中,无人机每隔一定距离记录一次位移信息、偏航角与图像关键帧;在回程时,无人机先根据位移信息到达关键帧附近,然后通过图像匹配纠正误差,对下一帧重复这两ki直到回到起点。仿真实验证明本文所述方法能实现自主返航功能,且可以应对复杂的路径与偏航角改变的情况。
[Abstract]:In recent years, the rotor UAV has been attached great importance for its excellent maneuverability. In order to make the control more convenient and accurate, the position and attitude control of UAV based on visual information has been a hot topic. Autonomous return is also an important function of UAV. At present, most of the flight control systems of multi-rotor UAV products provide one-click return function based on satellite navigation signal, but the satellite navigation signal is not reliable in some complex environments. The purpose of this paper is to accurately control the position and attitude of the rotoring UAV based on visual information, and to realize the function of autonomous return. Firstly, this paper introduces a speed controller to improve the performance of UAV position and attitude control based on the fusion of optical flow and inertial navigation devices of the camera image at the bottom of the UAV. The flight experiments of the four-rotorcraft show that the proposed control algorithm can better achieve the land-based hovering of fixed points, independent tracking, and the integration of velocity information can obtain a more accurate relative displacement of UAV. On this basis, this paper uses the UAV front camera to achieve autonomous return function. According to the theory of multi-view geometry of computer vision, the 3D transformation relationship between the two angles of view can be solved by matching the current camera image with the key frame, thus controlling the UAV to approach the position of the key frame and hover. The robust autonomous return scheme proposed in this paper combines the estimation of velocity navigation with image matching hovering: during the journey, the UAV records displacement information, yaw angle and image key frame at a certain distance. The UAV first arrives near the key frame according to the displacement information, then corrects the error by image matching, repeats the two Ki to the next frame until it returns to the starting point. The simulation results show that the proposed method can realize the autonomous return function and can deal with the complex path and yaw angle changes.
【学位授予单位】：上海交通大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：V279;V249

【参考文献】