• CN:11-2187/TH
  • ISSN:0577-6686

机械工程学报 ›› 2019, Vol. 55 ›› Issue (14): 169-177.doi: 10.3901/JME.2019.14.169

• 交叉与前沿 • 上一篇    下一篇

复杂异形曲面自动建模图像矩视觉伺服定位控制

彭伟星1,2, 王耀南1,2, 曾凯1,2, 吴昊天1,2   

  1. 1. 湖南大学电气与信息工程学院 长沙 410082;
    2. 湖南大学机器人视觉感知与控制技术国家工程实验室 长沙 410082
  • 收稿日期:2018-12-07 修回日期:2019-05-07 出版日期:2019-07-20 发布日期:2019-07-20
  • 通讯作者: 王耀南(通信作者),男,1957年出生,博士,教授,博士研究生导师。主要研究方向为智能控制理论与智能信息处理。E-mail:yaonan@hnu.edu.cn
  • 作者简介:彭伟星,男,1994年出生,博士研究生。主要研究方向为机器人自主三维建模视觉伺服控制。E-mail:18711047309@163.com
  • 基金资助:
    国家自然科学基金资助项目(61733004,61573134,61433016)。

Positioning of Autonomous Modeling Manipulator for Deformed Surface Using Moments-based Visual Servoing

PENG Weixing1,2, WANG Yaonan1,2, ZENG Kai1,2, WU Haotian1,2   

  1. 1. College of Electrical and Information Engineering, Hunan University, Changsha 410082;
    2. National Engineering Laboratory for Robot Visual Perception and Control Technology, Hunan University, Changsha 410082
  • Received:2018-12-07 Revised:2019-05-07 Online:2019-07-20 Published:2019-07-20

摘要: 异形曲面目标的表面深度信息无法进行参数化表示,是难以建立矩特征交互矩阵的主要原因。提出一种利用目标轮廓深度信息直接构建矩特征交互矩阵的新方法,推导出传统图像矩交互矩阵的非参数化形式。考虑到传统矩特征视觉伺服定位过程的响应速度慢、收敛范围小,设计一种新型的深度矩作为视觉伺服的图像特征,并推导出其非参数化交互矩阵。采用最速下降法设计视觉伺服控制器,分析并比较两种矩特征视觉伺服控制方法在异形曲面定位过程的动态性能。结果表明,非参数化形式的传统矩交互矩阵和深度矩交互矩阵均有效,两种控制方法在初始偏差较小时是渐近稳定的。提出的深度矩视觉伺服方法与传统的矩特征视觉伺服相比,较为明显地改善了收敛速度和收敛范围。

关键词: 定位控制, 复杂异形曲面, 图像矩视觉伺服, 自动建模

Abstract: The surface depth information of the deformed surface target cannot be parameterized, which is the main reason why it is difficult to establish the image interaction matrix of moment feature. Therefore, a new method for constructing the nonparametric interaction matrix of the traditional image moment is derived based on the target contour depth information. Considering that the traditional moment based visual servo control has slow response speed and small convergence range, a novel depth moment is designed as the visual servo image feature, and its non-parametric interaction matrix is derived. The visual servo controller is designed using the steepest descent method, and dynamic performance of the two moment feature based visual servo control methods in the process of profiled surface positioning is analyzed and compared. The results show that nonparametric form of both the traditional moment interaction matrix and the depth moment interaction matrix are effective. Both methods are asymptotically stable in the control process. Compared with the traditional moment feature based visual servo method, the depth moment based visual servo improves the convergence speed and convergence range.

Key words: autonomous modeling, deformed surface, moments-based visual servoing, positioning

中图分类号: