• CN:11-2187/TH
  • ISSN:0577-6686

机械工程学报 ›› 2019, Vol. 55 ›› Issue (3): 64-72.doi: 10.3901/JME.2019.03.064

• 机构学及机器人 • 上一篇    下一篇


訾斌, 尹泽强, 李永昌, 赵涛   

  1. 合肥工业大学机械工程学院 合肥 230009
  • 收稿日期:2018-06-27 修回日期:2018-10-31 出版日期:2019-02-05 发布日期:2019-02-05
  • 作者简介:訾斌,男,1975年出生,博士,教授。主要研究方向为机电控制及自动化。E-mail:binzi.cumt@163.com;尹泽强,男,1993年出生,男,硕士研究生。主要研究方向为柔索并联机器人动力学与控制。E-mail:xiaoqiang22ll@sina.com
  • 基金资助:

Fast Mobile Component Location Method for Cable-driven Parallel Robots Based on YOLO Model

ZI Bin, YIN Zeqiang, LI Yongchang, ZHAO Tao   

  1. School of Mechanical Engineering, Hefei University of Technology, Hefei 230009
  • Received:2018-06-27 Revised:2018-10-31 Online:2019-02-05 Published:2019-02-05

摘要: 针对柔索并联机器人移动构件实时定位问题,提出一种基于YOLO目标检测模型的柔索并联机器人移动构件快速定位方法。首先根据YOLO目标检测模型设计深度卷积神经网络结构,根据PASCAL VOC数据格式构建自己的数据集,并在该数据集上训练及测试模型,然后将工业摄像机采集得到的图像数据输入模型中进行标靶检测,记录标靶的类别和位置。分析标靶的颜色特征,并将标靶图像进行二值化,进一步计算出柔索并联机器人的精确位置。试验表明该方法能对图像中的目标进行准确分类和定位,定位误差在1°以内,图片处理帧率可达33帧,满足实时性要求,同时算法具有良好的准确性和有效性。

关键词: 快速定位方法, 目标检测, 柔索并联机器人, 深度学习

Abstract: Aiming at the problem of real-time mobile component localization of cable-driven parallel robot, a fast mobile component location method based on YOLO object detection model is proposed. Firstly, the deep convolution neural network structure is designed, a data set is constructed in PASCAL VOC format, and the model is trained and tested on that data set. Then the image data collected by the industrial camera are inputted into the model to detect the target, record the category and position of the target, analyze the color characteristics of the target and binarization. The precise position of the mobile component of the cable-driven parallel robot is further calculated. The experimental results show that the method can classify and locate the target accurately. The positioning error is less than 1°, and the frame rate of image processing can reach 33 frames, which can meet the real-time requirements. At the same time, the algorithm has good accuracy and effectiveness.

Key words: cable-driven parallel robot, deep learning, fast location method, object detection