结合深度学习的机器人示教系统设计

传统的机器人示教系统在使用上受到应用对象和硬件设备的限制，导致其开放性和易用性较低，为了降低示教系统的使用门槛，提高人机交互的效率，利用ROS（Robot Operating System）的开放性和跨平台性，设计了手势引导机器人示教系统，可以控制机器人进入学习、编码、执行等模式。系统采用YCbCr与RGB空间相结合的肤色分割算法，利用CNN深度学习框架进行特征提取完成手势识别；基于ROS集成手势对机器人模式控制。通过在公开数据集上实验验证手势识别准确率可达96.49%，并测试了系统的有效性与可靠性。

Design of robot teaching system combined with deep learning

The traditional robot teaching system is limited by the application objects and hardware equipment, which leads to its low openness and ease of use. In order to reduce the threshold of the teaching system and improve the efficiency of human-computer interaction, ROS (Robot Operating System) is open and cross-platform. A gesture-guided robot teaching system is designed to control the robot to enter learning, coding, and execution modes. The system uses a skin color segmentation algorithm combining YCbCr and RGB space, and uses CNN deep learning framework for feature extraction to complete gesture recognition; ROS integrated gesture control for robot mode. Experiments on public data sets verify that the accuracy of gesture recognition can reach 96.49%, and the effectiveness and reliability of the system are tested.