基于单目视觉伺服系统的高速旋转球体三维速度测定

3D Velocity Measurement of High-speed Rotating Sphere based on the Monocular Vision Servo System

单目视觉所处理的信息是目标物体在三维空间内的信息向二维平面的投影,故还原目标的深度信息,对于预测目标轨迹和后续控制决策具有重要的意义。提出一种基于单目视觉的高速旋转球体三维速度的方向及大小的测定算法,利用球体上单个标记的运动轨迹来识别球体旋转速度及旋转轴的方向。摄像机角度固定,当球体旋转轴正对摄像机时,球体标记运动形成的运动轨迹是一个圆,而当球体旋转轴方向发生改变时,球体标记运动轨迹将会变成一个椭圆,不同的方向形成椭圆的长短轴不一样,利用这个特征来计算旋转轴的方向。该算法可用于测量乒乓球旋转中,旨在应用于乒乓球机器人的视觉识别,对机器人后续阶段的运动规划和控制决策起到重要的作用。对所提算法进行了实验验证,得到与理论结果误差较小的实验数据,验证了方法的可行性。

The information monocular visual processed is a target object that project from 3D to 2D. So restoring the depth information is very important to the prediction of target trajectory and subsequent control decision. This paper presents an algorithm based on monocular vision that can calculate 3D velocity of high-speed rotating sphere. The algorithm mainly uses the single trajectory on the sphere to identify the rotation speed of the sphere and the direction of the rotation axis. The angle of the camera is fixed, and when the sphere rotates on the camera, the moving track formed by the ball mark motion is a circle. When the direction of the axis of rotation of the sphere is changed, the trajectory of the mark on the ball will become an ellipse. When the direction of the rotation axis is different, the shape of ellipse becomes different. This algorithm can be used to calculate the rotating velocity of table tennis ball. The algorithm will be applied to vision recognition of table tennis robot, which play an important role in motion planning and control decision of robot. The experimental results of the proposed algorithm are verified with little error, which proves the feasibility of the proposed method.