Information Technology Reference
In-Depth Information
Chapter 9
Image Measurement Errors in Visual Servoing:
Estimating the Induced Positioning Error
Graziano Chesi, Yeung Sam Hung, and Ho Lam Yung
Abstract.
The goal of a visual servo system is to position a robot end-effector by
progressively adjusting its location so that some object features in the current im-
age match the same features in a desired image previously recorded. However, this
matching in the image domain cannot be ensured due to unavoidable presence of
image measurement errors, and even when it is realized, there is no guarantee that
the robot end-effector has reached the desired location since the available image
measurements are corrupted. The aim of this chapter is to present a strategy for
bounding the worst-case robot positioning error introduced by image measurement
errors. In particular, two methods are described, which allow one to compute upper
and lower bounds of this positioning error. Some examples illustrate the proposed
methods with synthetic and real data.
9.1
Introduction
The teaching-by-showing approach is a fundamental technique for eye-in-hand vi-
sual servoing. In this approach, a robot end-effector mounting a camera is positioned
at a location of interest, called desired location, from which some object features are
visible. The view of the camera at this location, called desired view, is hence stored.
Then, the camera is moved to another location of the scene from which the same
object features are visible. The target is to steer the camera from its current location
to the desired location by exploiting as feedback information the view of the camera
in the current location, called current view, and the desired view previously stored.
Various methods have been proposed in the literature for the teaching-by-showing
approach, see for instance [21, 19, 4, 5]. Among these, there are basically two
pioneering methods: image-based visual servoing (IBVS), see
e.g.
[20, 3], and
Search WWH ::
Custom Search