基于双目视觉的主从式机器人主手系统设计和研究

提出了一种基于双目视觉的主从式机器人主手系统，采用双目视觉技术实时识别并跟踪手柄的运动，同时测量其空间位姿。首先通过背景差分法检测出手柄，然后用卡尔曼滤波器对手柄的位置和速度进行预测，跟踪手柄的运动。在标记点的立体匹配中采用了特征匹配和区域灰度匹配相结合的方法，并引入唯一性和顺序一致性约束来剔除伪匹配，最终测罱出手柄的实时空间位姿信息。实验结果证明系统方案是可行的。

The Master Manipulator System Based on Binocular Vision for Master-slave Robot

A master manipulator system based on binocular vision is proposed for the master-slave robot. The system employs binocular vision technology to identify and track the movement of the manipulator in real-time, and also measures the position and attitude. Firstly identify the manipulator by background difference method, then use Kalman filter to predict the position and velocity of the manipulator for tracking the movement. The method that combines feature matching with local gray scale matching is employed to the stereo matching of marking point and the constraint of uniqueness and sequence consistency is used in excluding false matches. The position and attitude of manipulator is calculated through measuring the three-dimension coordinate of marking points. Finally, experiment result validates the method proposed in this paper.