基于立体视觉的无人机动态目标实时跟踪

Real-time tracking of UAV dynamic target based on stereo vision

针对小型四旋翼飞行器的动态目标跟踪问题,提出了连续自适应椭圆检测(Continuously Adaptive Ellipse Detector,CAED)算法的飞行器目标定位以及一种基于立体视觉的跟踪控制算法。首先,双目相机拍摄的图像通过CAED算法得到目标的像素中心点,利用三角测量原理,实现了对目标位置的有效估计。其次,针对运动目标跟踪任务,设计并实现了两个级联串级PID的四旋翼位置姿态控制律。最后,在四旋翼上将图像处理算法与四旋翼控制相结合,进行了测试实验,验证了本文所提出的检测和控制方法。

Aiming at the dynamic target tracking problem of small quadrotors,a continuous adaptive ellipse detection(Continuously Adaptive Ellipse Detector,CAED)algorithm for vehicle target positioning and a tracking control algorithm based on stereo vision are proposed.First of all,the image taken by the binocular camera obtains the pixel center point of the target through the CAED algorithm,and the effective estimation of the target position is realized by the principle of triangulation.Secondly,for the task of moving target tracking,two cascade PID four-rotor position and attitude control laws are designed and implemented.Finally,the image processing algorithm is combined with the quadrotor control on the quad-rotor,and test experiments are carried out to verify the detection and control method proposed in this paper.