臺灣博碩士論文加值系統

本篇論文旨在整合攝影機及機器手臂，成為一具有視覺回授(visual
feedback)的視覺伺服 系統(visual servoing system)，再利用此伺服
系統來作物體追蹤。視覺伺服系統大致上分成位置基準控制(position-
based control)及影像基準控制(image-based control)，本文 採取位置
基準控制作為整個系統的架構。為達成此系統建構，將此系統細分成三部
分。影像部分採用減去背景的方法找出物體，並用重心法求得物體之位置
。路徑預測部分則採用卡門濾波器(Kalman filter)，以對物體的移動路
徑進行預測。機器手臂控制部分主要以梯形法 來計算機器手臂的移動路
徑。此外利用神經網路控制器來消除回授訊號中，因不良的環境而產生的
雜訊。

A visual servoing system, which is consisted of a CCD camera
and a robot, is presented in this paper. Through the visual
feedback from the CCD camera, the proposed system is able to
track a moving object. In gereral, the visual servoing systems
can be approximately divided into the position-based control and
the image-based control respectively. Theposition-based control
method is adopted in the proposed system. The system contains
three major elements: image process, trajectory prediction,and
robot control. In image process, a method of reducing the
background effect is employed to find out the object, and then
thecenter of the gravity of the object image is the object
position. In trajectory prediction, the Kalman filter is
equipped to predict the trajectory of themoving object. In robot
control, a ramp velocity profile is used to regulate theposition
command for controlling the manipulator. Additionally, a neural
network controller is adopted to eliminate the noise imposed on
the feedback signal, if the system is under a coarse enviroment.