Use of Machine Vision in Autnomous Robotics
Erik Blasch
Mechanical Engineering Dept.
University of Wisconsin-Madison
2:30 pm Fri. Apr. 19 in 2310 CS&S
Machine vision, real-time use of vision sensing, has proved successful
in part inspection and identification in manufacturing and is showing
promising results for applications in autonomus robotics. The purpose
of the presentation is to give the audience a perspective of the
issues involved and how the presenter has implemented five
machine-vision sensing systems for autonomous robots. The
presentation is divided into three sections, 1) computer vision versus
machine vision, 2) examples of machine vision systems for autonomous
robotics, and 3) current research in intelligent sensor fusion of
machine vision for autonomous robotics.
Vision sensing in robotics needs to be fast and accurate as the robot
typicallly moves in a dynamic environment. Given these constraints on
performance, autonomous robot machine vision systems require 1) a well
defined objective, 2) acclimation to the environment of interest, and
3) a strategic algorithmic implementation plan. With known
objectives, such as position sensing, target identification, or
obstacle avoidance; implementation of machine vision sensing for
robotics is realizable. Likewise, a structured environment,
(i.e. manufactuing center) with known landmarks or fiducials has
proved favorable. In addition to the above reasons; by combining vision
data with other sensory data, the machine vision limitations on update
rates are minimized for successful real-time task performance.