双目定位多功能机械臂系统设计

Design of Binocular Positioning Multifunctional Manipulator System

摘要: 准确性、快速性与鲁棒性是机械臂系统设计的要求，经过研究提出一套新形式的机器人视觉系统，利用非接触式测量，对所操控单元进行位姿计算，完成操控单元的三维信息与机械臂执行单元进行命令交互。设计采用双目CMOS工业相机，通过Matlab进行相机标定并得到内外参数，然后进行立体校正，结合SGBM立体匹配算法解算出位置深度信息，进而对应帧的二维图像，利用OpenCV视觉算法获取操控器下端标志物的三维坐标，传递执行指令给机械部控制单元。由实验结果分析，在白光环境下该机器人系统同步现实功能展现出良好的跟随准确性，精度误差始终保持在4.5%以内，命令响应时间短，鲁棒性好。

Abstract: Accuracy, rapidity and robustness are the requirements of the design of the manipulator system. After research, a new form of robot vision system is proposed, which uses non-contact measurement to calculate the position and posture of the controlled unit, and completes the command interaction between the three-dimensional information of the control unit and the manipulator execution unit. The design adopts a binocular CMOS industrial camera. The camera is calibrated and the internal and external parameters are obtained through Matlab, and then the stereo correction is carried out. The position depth information is calculated by combining the SGBM stereo matching algorithm, and then the corresponding two-dimensional image of the frame is obtained. The OpenCV vision algorithm is used to obtain the three-dimensional coordinates of the lower marker of the manipulator, and the execution instructions are transmitted to the control unit of the mechanical department. According to the analysis of the experimental results, the synchronous reality function of the robot system shows good tracking accuracy in the white light environment, the accuracy error is always kept within 4.5%, the command response time is short, and the robustness is good.