基于图像矩的机器人视觉伺服  

Visual Servoing Using Image Moments

作　　者：林靖[1] 陈辉堂[1] 王月娟[1] 蒋平[1] 

机构地区：[1]同济大学电气工程系,上海200092

出　　处：《同济大学学报（自然科学版）》2000年第4期438-442,共5页Journal of Tongji University:Natural Science

基　　金：高等学校博士学科点专项科研基金!(9424713)

摘　　要：区别于图像的简单几何特征，利用图像的全局特征描述子──图像矩特征作为图像特征信息，推导了矩特征变化量与相对位姿变化量之间的关系矩阵，即图像雅可比矩阵．针对平面视觉伺服控制，利用了所推导的基于矩特征的图像雅可比矩阵，在不需要知道目标的成像高度、摄像机焦距及不需要精确标定的摄像机外部参数的情况下实现了基于图像的视觉伺服控制．最后给出了仿真与实验例子，表明了矩特征的有效性．This paper describes using image moments as image features for visual servoing. Unlike the simple features such as points, lines or circles, image mements are the global image descriptors. Using image moments, the complex feature matching process is avoided, and making the special marks in the scene is unnecessary. With the Moments- based Jacobian which is deduced in this paper, the planar visual servoing is realized with unknown depth of object and camera focus and also inaccurate external parameters of camera. The simulation and experiment results show that image mements can be used as image features for visual servoing.

关 键 词：视觉伺服 机器人 图像矩 雅可比矩阵 

分 类 号：TP242[自动化与计算机技术—检测技术与自动化装置]