一种基于视觉识别的按钮自主操作机械手

Autonomous manipulator for button pressing based on visual recognition

针对机器人自主按压按钮的操作需求,设计了一种适于机器人安装的基于视觉识别的四自由度(Degree of freedom,DOF)按钮自主操作机械手。结合功能需求对该机械手的机械结构和控制系统进行设计,通过几何法求解了机械手的运动学正反解,并规划了其运动轨迹。采用支持向量机对按钮样本进行训练和预测,并通过视觉对按钮进行定位。最后制作了按钮自主操作机械手样机,并进行了按钮识别、机械手定位和按压按钮实验。实验结果表明,该文4-DOF自主操作机械手的按钮识别成功率达94%,控制定位精度达10 mm,能够自主完成按钮按压操作。

According to requirements of the autonomous button pressing of the robot, the four-degree of freedom（4-DOF） autonomous manipulator for button pressing of the robot is developed based on the visual recognition. The mechanical structure and control system of the manipulator are designed according to its functional requirements, the kinematic is solved by using the geometric method, and the movement trajectory is planned. The support vector machine （ SVM） is used to train and predict the image samples of the buttons, and the buttons are positioned visually. Lastly, a prototype of the manipulator is developed. The experiment results of button recognition, manipulator positioning and button pressing show that, the success rate of the button recognition of the manipulator sample is94% ,the positioning accuracy of the manipulator control system is 10 mm,and the developed 4-DOF manipulator can press the target button autonomously.