基于机器视觉的桁架机器人自动上下料系统的设计
Design of Automatic Loading and Unloading System for Truss Robot Based on Machine Vision

针对电力系统电缆保护管自动垂直仓储问题，研制了基于机器视觉的桁架机器人自动上下料系统。首先设计了由桁架机器人本体、视觉导引系统、机器人运动控制系统三部分组成的自动上下料系统，介绍了机器人本体结构，系统控制策略；其次在搭建了系统的手眼定位系统的基础上，详细介绍了深度可分类卷积神经网络分类识别模型；最后通过系统的现场测试和验证。实验结果表明深度可分离卷积神经网络识别模型的符合设计预期，实现了待测物体的图像分割，分类和定位。桁架机器人自动上下料系统，符合工程应用要求，实现了高效，稳定的自动上下料功能。
According to the automatic vertical storage problem of the power system cable protection pipe, the automatic feeding system of the truss robot based on machine vision is developed. Firstly, the automatic loading and unloading system composed of the truss robot body, the robot guidance system and the robot motion control system is introduced. Secondly, the model is introduced. Finally, the field test and verification of the system are passed. The experimental results of the clear and deep separable convolutional neural network recognition model meet the design expectation, and realizes the image segmentation, classification and positioning of the objects to be measured. The truss robot automatic loading and unloading system meets the requirements of engineering application and realizes the efficient and stable automatic loading and unloading function.